国外智能网联汽车综述

【实例简介】
最新国外大牛写的智能网联汽车综述，车辆工程，伯克利大学
5 3 limI 6 5 4 2 Figure 1: CarToon depicting a variety of intelligent transporTation system on highway, arterial and urban roads enabled by connected and automaled vehicles(CAvs) Each number refers to a CAV application discussed next. Communication with other vehicles enables (1)augmented awareness, (2) platooning, and (3 )cooperative maneuvers. Communication with the infrastructure enables(4)enhanced approach and departure to signalized intersections. Cloud connectivity enables access to databases forec nd remote computations. On-board perception, localization and maps are fundamental to navigate in known and unknown environments, that all include nunl-cunnected vehicles, cyclists, pedestrians. In(5) roadway sensors generate signal phase and lining(SPaT)and vehicle occupancy and speed (VOs) data,thatcanbestoredinthecloudoTherapplicationsincludecoordinationofgridchargingparkingroadworks(6).(createdonhttps://icograms.com and parking. The dSRC technology may also be used to im- In any powertrain, auxiliary loads(e.g. air conditioning prove GPS accuracy [37 and for geo-fencing区域限定 lights, infotainment) can significantly affect the overall en Internet connectivity via cellular communication enables ac- ergy consumption. For systems that are not safety-critical, the cess to cloud-based data and services. Due to its low latency, level of performance may be temporarily decreased to limit the 5G may also compete with dsrC for V2V and V2I communi- power consumption cations Advantages of the DSRC technology include security, low la- 2.2. Infrastructure tency, interoperability, and resilience to extreme weather condi The roads on which CAVs are operated include complex sys tions; on the flip side, it requires dedicated hardware. 5G offers tems for traffic monitoring and control. Static and dynamic both access to multimedia and cloud services(that are highly maps databases, and remote computations can be accessed by valued by customers), and cooperation with other vehicles and CAVs through the cloud infrastructure. The solution in the future may be a combination of both technologies 38 Highway infrastructure 环形探测器、磁探测器、磁力仪」 Modern highways are instrumented with systems for vehi Vehicle powertrains cle detection, to monitor and potentially control the traffic flow By controlling vehicle motion with increased awareness, A variety of detection technologies are employed, including CAVS can inherently improve energy usage. Additionally, pow- in-roadway sensors (loop detectors, magnetic detectors, mag entrain control systems can benefit from the forecasts that maps, netometers)and over-roadway sensors(cameras, radars, ultra perception, and communication make available sonic, infrared, and acoustic sensors)[40]. The uses of sensors The majority of today's powertrains are based on internal in highways include data collection(vehicle occupancy, speed, combustion engines; more advanced powertrains(sometimes type) for monitoring and planning of road use [411, and active called micro- and mild-hybrids)can include start/stop systems, traffic control via ramp metering engine coasting systems, and some energy regeneration [39 道控制 Hybrid electric vehicles(HEVs) have high voltage, medium Urban infrastructure capacity battery packs, an electric motor, and an internal com- A typical instrumented intersection(see e. g. [42)features bustion engine, allowing pure electric driving, pure thermal in-roadway sensors, like loop detectors or magnetic sensors driving, and hybrid driving. In plug-in HEVs, the battery can be that detect the presence of vehicles at a stop bar. Additional echarged from the grid and the battery pack is typically larger sensors at advance locations and in departure lanes can also be than in HEVs, allowing to drive on electricity for significant used to estimate the vehicles speed and turn movements [43] distances The signal phase and timing(referred to as SPat and describin The appeal of purely electric vehicles is due to the absence the current light color and the remaining until the next change of local emissions, low price of electric energy, good dynamic of color) can be retrieved directly from the controller, or indi- performance and low noise rectly via image analysis The uses of these data include the analysis of intersection Remote Planning and Routing performance, tuning of controllers, feedback to adaptive con- trollers and broadcasting to vehicles for coordination and traf Eco-Rouling Ma fic flow improvement [44]. Controllers can implement a fixed cycle, change green times depending on immediate traffic con- Eco-Driving and coordination Historical data ditions, or implement more advanced control strategies adapt- ing to congestion level; pedestrians can be part of the cycle Battery Charge or make requests with buttons. Metrics for intersection per Planning formance include volume-to-capacity ratios, fraction of arrivals 一一 in green, red-light violations, queue delays [45 46]. These data Real-Time Control and Planning may be stored locally and collected by operators, or stored re motely on the cloud Motion Planning Intersections can be instrumented to broadcast messages to Sensor fusion (Map-based) Localization nearby vehicles using DSRC. The SAE standard 2735 [28] in- Motion cludes messages for signal phase and timing and intersection Control Environment prediction geometry. Notice that the timing part is deterministic only if Powertrain the controller has a fixed cycle; otherwise, the timing is inher- Control ently uncertain because of the stochastic nature of traffic Beyond intersections and traffic signals, urban infrastructure includes fuel stations, charging stations, and parking infrastruc ture. A connected charging and parking infrastructure enable better routing of vehicles and more effective pricing schemes Vehicle charging has a significant effect on grid balancing and smart grids [47 4849]. Automated parking systems enable Figure 2: Architecture for Connected and Automated Vehicles(CAvs) deploy better exploitation of urban surfaces. ment Cloud infrastructure Cloud services supply to CAVs static and dynamic road more complex protocols [55 for which a standard has not yet maps,historical databases, and remote computational power been established. An appealing niche for cooperative vehicles Access to the cloud is enabled by cellular connections is freight transportation; heavy duty vehicles capable of auto Modern road and traffic map services(see e.g. 150 5152)) mated driving and v2V communication can form platoons and provide information that goes beyond the maps for navigation, drive at small inter-vehicular distance, thereby reducing their including air drag resistance and fuel consumption(see e. g. 1561) static information, such as road grade, road curvature, lo- Non-cooperative vehicles, cyclists and pedestrians cation of intersections, lane maps, speed limits, location When interacting with non-cooperative road users, CAVs do 交叉口延误 fuel and charging stations, intersection average delays not differ substantially from other self-driving vehicles. In dynamic information, such as traffic speed availability and this case, awareness of the surroundings relies on the percep price of fuel and charging stations, intersection del tion system. This includes non-cooperative vehicles, cyclists traffic congestion, road works, weather conditions pedestrians, and any other road user. Recent research and tech nologies are oriented to some level of cooperation with cyclists Historical data can be relevant for planning problems, like and pedestrians, enabling safety communications between ve vehicle routing and reference velocity generation. Examples in- hicle and smartphones [33571 clude traffic congestion on highways 41, and signal phase and timing data [5354); in both cases, historical data give deeper 3. Connected and Automated Vehicle Control Architecture insight on traffic patterns, that is generally not found in maps CAVs can perform computations remotely, using cloud ser- Figure shows a control architecture for Connected and Au- quirements. Computations moved to the cloud may include operation. The architecture includes on-board and remote func (dynamic) routing and long-term trajectory optimization tional block 23. Other road users 迹优化」 3.I. Real-time control and planning Other connected and automated vehicles The functional blocks that reside on-board are safety-critical Multi-vehicle cooperation happens among two or more and need to be executed in real-time. The real-time layer inter CAVs. The aforementioned SAE J2735 standard [28 is mostly faces to the vehicle actuators, collects measurements from on- oriented to awareness and safety applications. Advanced ve- board sensors, and performs all the real-time computations that hicle cooperation, including multi-vehicle formations, require make a CAv reliable and robust to unpredicted events. These computations include the control and planning algorithms that Eco-driving and coordination are shortly described next, and detailed in Section 4 The eco-driving and coordination block takes route informa tion and computes a reference velocity trajectory for the on board algorithms. The value of this block is in the use of long Powertrain control term forecasts (like road grade and traffic congestion) and in the Powertrain control depends on the powertrain type and may accounting for constraints like trip time and maximum velocity nclude engine control, electric motor control, gear shifting Some constraints depend on the driving context: for instance control. Powertrain controls satisfy in real-time the power re passing a signalized intersection during a green phase; histori quired to move the vehicle, and affect the so-called"tank-to wheel"energy conversion [39]. Reactive controls select cal data may help to improve performance In these driving scenarios, the ego-CAV can cooperate with the powertrain operating points based on the current power de- other CAVs. An example of multi-vehicle coordination is pla- mand. Energy efficiency can be improved when forecasts are mooning, in which a group of vehicles travel on a certain road available, both for the short-term(speed and torque profiles segment at reduced distance gaps [2]. The objective can be from longitudinal control)and long-term(from the cloud layer). to maximize the usage of road surface (and hence throughput) or to reduce the aerodynamic drag. In the multi-vehicle case. Motion control the eco-driving block uses the same information, but the prob The motion control block regulates the longitudinal and lat- lem is generally more complex eral motion of the vehicle, and is interfaced to the powertrain 气动力阻力 controls and the steering system. The desired vehicle motion Eco-routing is generally specified at a higher hierarchical control level, and The eco-routing block determines the most energy-efficient the motion control ensures that the reference behavior is exe route, given user requirements and road maps(e. g road grade cuted in closed loop. Motion control affects safety and the so- traffic speed, intersection delays, fuel or charging stations) This block outputs the optimal route, i.e. a set of way points called"wheel-to-distance"energy conversion(3958 When along with the intersection locations, speed limits, road grade forecasts of trafic, signals, and trajectories of other vehicles are available, safety and performance can be significantly im- proved 33. what is not covered in this survey The real-time planning and control blocks require feedback 策略规划 from the vehicle, its position and velocity relativ Motion planning rounding environment, and predictions of moving obstacles The real-time planning block includes maneuver planning [61 A CAV may be equipped with a GPS unit for localiza- (e.g. decision to stay in a lane or change), path planning, and tion, cameras, radars and lidars for perception and a dSRo trajectory planning. These blocks also depend on the driving unit for V2V and V2I communication. These data are processed context, and their boundaries are quite blurred (591 and fused to estimate the position and velocity of the CAv and the surrounding objects, both static and moving. To cope with agents like pedestrians, cyclists and non-connected vehicles, an 移动 3. 2. Remote planning and routing algorithm predicts the future trajectories of moving obstacles. BETE The remote layer in Figure 2]enables access to external data Perception, localization, and environment prediction are ex 物 tremely important for self-driving vehicles and CAvs. The in- sources, and performs longer term computations, that mostly terest of the academic and industrial research communities on affect performance and are not real-time critical. These com- these topics is very high and has produced a vast literature putations include the planning and routing algorithms that are shortly described next, and discussed into details in Section To limit the scope of this survey, we only focus on the real following a bottom-up order time control and planning layer and on the remole planning and mmg线层 Battery charge planning匾电式混合动力电动 3.4. How to read this survey If the Cav is an electric, hybrid, or plug-in hybrid electric ve In the next two sections, we will analyze the functional hicle,a long-term planning of the battery charge trajectory can blocks in the real-time control and planning layer(in Section 4) prevent suboptimal utilization of the energy stored on-board In and the remote planning and routing layer(in Section 5), fol an electric vehicle, this algorithm can simply predict the driv- lowing a bottom-up approach. The actual inputs and outputs ing range using route information; if the range allowed by the of each block will be more precisely specified, improving the current battery charge is exceeded, the algorithm may alert the understanding of the overall architecture. For each block, our user, request to re-plan the route, or plan a stop in a charging main goal is to survey the existing literature. To limit the scope station In hy brid and plug-in hybrid electric vehicles, an inter- of this survey, we mostly focus on optimization-based methods nal combustion engine is available; the route information can be and energy efficiency, and we point to more focused surveys on used to optimize the allocation of fuel power and battery power specific topics. Contextually, we highlight the challenges and along the trip opportunities enabled by CAVs. Opportunities are often related 根据上下文 Vehicle type Fuel Electric (Plug-in) W powered Hy brid E T Gear shifting B Engine on/off (a) Intcrnal combustion cnginc vchicle Energy management (b) Electric vehicle. Table 1: The powertrain control probleMs surveyed in this paper, and their E applicability to the most common powertrains B to automated and cooperative driving, improved environment forecasts, and connectivity for data and remote computations. (c)Pre-Lranisinlissionl or single-shall parallel hybrid electric vehicle 扑结构 Driver safety, performance improvement, and real-time opera- Figure 3: Commnon powertrain topologies. Thin lines: electrical connections. tion are identified as the main technical challenges, real-time Thick lines: mechanical connections. W: longitudinal dynamics. T: transmis- operation includes the coordination between on-board and re- sIon. E: internal combustion engine. M: electric motor. B: high-voltage battery mote layers. Where pertinent, we illustrate selected approaches C: clutch. Further powertrain topologies, including series and combined hy brids,are presented in [39 with more detailed examples On-board real-time control and planning where T is the duration of the driving schedule, P, is the power extracted from the fuel, Pa is the battery internal power, yr and In this section, we review the existing literature for each of ya are their weights. It is easy to determine the optimal policy the three functional blocks in the real-time control and plan- for a fixed profile of the power demand; for instance, the opti ning layer of Figure 2 powertrain control, motion control, mo- mal gear shifting policy during a standard driving cycle may be tion planning. Performance metrics include vehicle energy con- computed by dynamic programming. In real-time operation the sumption passenger comfort, and-at a broader level-road power demand is not known in advance, but the optimal pol throughput. The main safety requirement is to avoid collisions icy can be approached by combining Model Predictive Control with other road users. This separation in blocks gives struc- (MPC) with accurate forecasts(e.g. of the power demand). We ture and facilitates the review. nonetheless the boundaries are now review some approaches that have been proposed in the sometimes blurred. Several of the works that we discuss inte-literature grate, at least partially, two or more of these blocks 41. Powertrain contro力总成控制 Gear shifting. Gear shifting control is available in automated transmissions, and impacts the way the upstream powertrain Powertrain control has a broad meaning and includes many components are operated in vehicles with manual transmis components and subproblems, such as transmissions, internal sion, it is commonly advised to"up-shift soon,, which trans combustion engines, electric motors, starters and generators At lates into operating the engine at low speed and high torque, large, powertrain controls address power generation for vehicle where efficiency is usually higher. We also know that this is motion and auxiliary loads. The literature on the topic is ex- possible only to some extent, because drivability(i.e.the re tremely vast; in this section, we focus on three powertrain con- sponsiveness of the vehicle to our inputs) is adversely affected trol problems in which connectivity and driving automation are Production gear shifting controllers are generally rule-based or can be leverage extensive testing and tuning can deliver good fuel economy and Literature review 换挡制~d If the future wheel speed and torque can be predicted reliably Different powertrain architectures allow more or less flexi- gear shifting control can be formalized as an optimal control bility in the realization of the power demand for vehicle motion problem and solved by various techniques. Since most trans- and auxiliary loads. In this paper, we focus on fuel-powered ve missions only feature a finite number of gears, the system dy hicles, electric vehicles, hybrid and plug-in hybrid vehicles(see namics are discrete. In[62 64 the gear shifting problem is FigureB3p. We survey gear-shifting control, engine on/off con- solved jointly with the energy management problem, combin trol,and energy management; Table I maps these three prob ing dynamic programming with Pontryagin's minimum princi lems to the different powertrain configurations. Gear-shifting and engine on/off are self explanatory. By energy(or power) ple in [62 and with convex optimization in [63]. In 16-4 also tt engine on/off is included; the resulting mixed integer non-linear management in hybrid vehicles, we refer to the problem of al 化 locating the power demand to the internal combustion engine program is treated as a distributed optimization problem, and4 formulated as a two-layer MPC problem. 65] uses the min and the electric motor imum principle and dynamic programming to jointly solve the In the three problems listed above, the goal is to minimize a gear shifting problem and the longitudinal control problem, for cost function of the form an fuel-powered vehicle (r Pr(+Yap,(dt, a simplified problem can be obtained assuming that the transmission gear ratio domain is continuous; in practice, this is only true when the vehicle is equipped with a continuously ECMS approaches for plug-in hybrids are summarized in 891 variable transmission. Even when this is not true, one can get A key aspect is the discharge rate of the battery; ideally, the bat- a suboptimal solution by rounding the optimal gear ratio to the tery is gradually discharged and reaches the minimum charge nearest available value 661 曝震控制」 only at the end of the trip. This requires route information and long-term planning, and is discussed in Section 5.1 Engine on/off. Engine control includes a vast family of chal- Some real-time approaches borrow the ECMS formulation; lenging problems, such as knock control, air/fuel ratio control, if the driving schedule is not known in advance, various up- thermal control(see e.g. 167 on the topic). Engine on/off con- date laws for A have been proposed, based on historical data trol determines whether to idle or shut fuel injection off. A and forecasts [9091. The generation of the reference state trivial approach to the problem is to cut injection as soon as the of charge, discussed in Section 5. 1 plays an important role power request is non-positive (in human-driven vehicles, when in this regard. Approaches that systematically address the in the gas pedal is released); this causes a sudden reduction of formation gap in real-time are mostly based on robust control torque and, ultimately, vibrations and discomfort. From a fuel [78 921, stochastic dynamic programming, and MPC Stochas economy perspective, restarting the engine has a cost, that is tic dynamic programming is used e. g. in 93] to minimize the generally lower than the cost of a cold start, but may be higher discounted infinite-horizon cost, and in 94 95 in a shortest than the cost of idling for a short time. Still, in favorable con- path formulation. A stochastic optimal control framework is ditions and with a sufficiently long preview of the upcoming developed in [9697 to determine the policy minimizing the driving profile, fuel savings between 5 and 10 were reported long-run expected average cost. All formulations yield a causal 6869 空转 time-invariant, state-feedback controller that can be fairly eas The engine on/off control problem is studied in [70] for a ily implemented conventional powertrain, using a hybrid systems formulation; MPC provides a systematic framework to include forecasts control design considers a relaxation to the continuous domain, and handle constraints in real-time. The authors of[9899 and maps the solution back into the hybrid domain. The same discuss a nonlinear MPC approach, in which an approximation problem is studied in [7172] for a belted starter alternator in a of the cost-to-go is derived using the relationship between dy hybrid electric vehicle, with the main focus being on vibration namic programming and Pontryagin's minimum principle. A and noise reduction. A similar setup has been considered in similar approach is proposed in [1001, where a preview of fu several other works, where the engine on/off and energy man- ture velocity is exploited. [101] is focused on the velocity pre agement problems are solved jointly [7363741; in this case, diction for MPC-based energy management. [102 proposes a the engine mode is often determined by dynamic programmin stochastic MPC approach, modeling the power demand from 随机MPC方法] the driver as a Markov chain and training it using standard driv Energy management. By energy management we refer to the ing cycles and historical driving data.[103[104 extend thi problem of allocating, in hybrid vehicles, the power demand approach showing how the driver model can be learned online to the internal combustion engine and the electric motor. This Although often disregarded in the scientific literature auxil problem has been extensively studied in the literature: we re- iary devices like air conditioning and lights can have a major fer the interested reader to 175, 76, 77 for extensive literature effect on energy consumption; to some extent, they can also be reviews and to [78 791 for systematic comparisons between ex- controlled. For example the air conditioning may be adjusted isting approaches to preserve the electric driving range| In an optimal control formulation, the limited energy stor- Instead of minimizing only energy consumption, several au age capability of batteries can be translated into a terminal state thors have addressed also different optimization goals, such as constraint, see e.g. 18081. In hybrid electric vehicles, the pointwise powertrain efficiency [1061, drivability [107 pollu- battery cannot be recharged from the grid, therefore the termi- tant emissions 63 1081, battery aging [109110, 1111, drivin nal battery charge is often constrained to its initial or nomi- cost [112,113114115]. The MPC approach in [1],in nal value. If the driving schedule is known in advance, this stead, combines longitudinal control and energy management problem is easily solved by dynamic programming [82,83]. exploiting forecasts of traffic signals and road slope The so-cal lled Equivalent Consumption Minimization Strategy (ECMS)can be derived from Pontryagin's minimum principle Challenges and opportunities for CAvs nd the observation that (under certain modeling assumptions) Gear shifting control, engine on/off control, and energy man- the adjoint state n(roughly speaking, the Lagrange multiplier agement are generally aimed at minimizing the cost function in associated to the terminal battery charge constraint) is constant for a fixed driving cycle; the optimal trajectory is found by it equation (Ip, and can benefit from forecasts of the vehicle speed and of the torque or power demand. These algorithms can nat- eratively determining the optimal d. The reader is referred to urally be integrated, to better manage the powertrain and the 184858687]for details on the model assumptions, guaran- associated uncertainty [64]. In CAVs this opportunity can be tees of optimality, implementation details, and performance in combined with more reliable forecasts. In facts, the future pre case the assumptions are violated files of vehicle velocity, wheel torque and power demand can In plug-in hybrid electric vehicles, the battery charge full be(to some extent) predicted, because of utilized, hence the trade off between fuel and electricity con sumption leads to an interesting optimization problem [88 driving automation and the removal (or substantial reduc tion)of unpredictable human factors we have implicitly assumed that T is a known nonlinear func- the awareness of the surrounding environment due to per tion of v and ng. The same can be said for the motor speed therefore, the motor power Pn is a known nonlinear function of ception sensors and communication with other vehicles v and Im. We wish to minimize the total powertrain energy and infrastructure In gear-shifting and engine on/off control, this opportunity &(x, u,w)=yr P/(ng, Se, Te, v)+Yg Pq(ng, Tm,v), mostly relates to the avoidance of energy-wasteful events: ev ery switching and shifting has a cost, and switching decisions where Pg=f(r, )-Eg and P is a nonlinear mapping from the are intrinsically reliant on forecasts or assumptions on the fu- engine speed and torque to the fuel thermal power; the mappin ture In energy management, we have documented how recent from v to the engine speed is implicitly embedded research has focused on filling the information gap on the future The input constraint set ul(v)=u(v)x12(v)xul x 14 defines the actuator limits where demand. Both short-term forecasts (as the ones just discussed) and long-term forecasts(which are handled as described in sec tion 5. Ip carry valuable information in this sense 1()={Tm:Tn(V)≤Tm≤Tm(v 7()={7:()≤Te≤T, Example: MPC approach for a plug-in hybrid electric vehicle In a plug-in electric vehicle, powertrain control includes gear 3={:l∈-1,0,+1)} shifting, engine on/off, and energy management. We formulate 14={ul:ll∈{-1,0,+1} it as the following finite-horizon optimal control problem in the time domain Tm, Tm, T, Te are nonlinear functions of the motor and engine velocities, and their mapping to the vehicle speed v is implicitly minimize g(xkr, Uxl, Wx )+l(xN) embedded. The state is confined to a safe operating region for MOr, LI t,.,I(N-ll the state of charge(to avoid overcharge or overdischarge), and subject to xk+t=f(xkt, ux,Wkt) to the discrete domains of the gear number and engine state 0=(x种,wp),}k=0,…,N-1, X=X1×X2×X3, where ∈7(wkt),xk∈X X1={EGn:E.≤Ea≤E =x,xMt∈X N 2 N N-I' be the solution at time (=4. The first x3={sc:se∈{0,1} input olt s applied, and at the next time step t=t+Tthe optimal control problem is solved using the new measurements The terminal battery charge must exceed a reference value rt. The mPC control law is ut=u XN=EG:Es<); in our architecture, E*is a position- We set the state vector to x=[La, ng, sel, the input vector dependent reference that is computed remotely in the charge [Tm, Te, ug, uelT, and the forecast vector to w=Iv, Pa, planning block( described in Section 5. 1p. In closed loop, this where Eg is the energy stored in the battery, ng is the gear num- constraint affects the actual battery charge at the end of the trip, ber, Se is the engine on-off state, Tn, is the motor torque, Te is which in turn affects the recharge time, i. e. the minimum wait the engine torque, ug is the gear shifting command, ue is the until the next trip. The terminal cost I is another knob that engine on/ off command, v is the vehicle longitudinal speed, Pa can improve closed loop performance in the long term; if it is the power consumption of electric auxiliaries approximates the optimal cost-to-go sufficiently well, it helps We model the powertrain dynamics as in[74] and we apply the mPC policy to approach the optimal infinite horizon pol- Euler discretization with step Ts, obtaining icy [117]. In this application, affine approximations of the form l=a+(eg-er )b have been shown to give good results [991 E TAle E2- 2Rs Q P,Eg We refer to [6264] for numerical techniques to solve prob (3) lem (2) and simulation analysis of the closed loop performance Se t ue 42. Motion control where Ab and Bh fit the battery open circuit voltage, R, is the Motion control ensures that the vehicle's longitudinal and lat battery internal resistance, O is the battery capacity The alg al motion follows a reference trajectory or path braic constraint h enforces the summation of Tm and Te at the longitudinal control is cruise control, which tracks a constant transmission imput shaft and the summation of motor n. pm reference velocity specified by the driver. Next we review the and auxiliary power Pa at the battery output main control systems for longitudinal and lateral motion h(x,u,w)sT (v, ng)-Tm-sety (4 Pb-Pn (v, Tm)Pa Literature review We first organize the existing longitudinal control approaches The input transmission torque T, is determined from a vehicle by their use of external information: predictive cruise control longitudinal model and from the transmission gear ratio; here (using a reference velocity computed remotely), adaptive cruise 8 control (adjusting the reference velocity based on the percep- predictive cruise control are also often proposed. More pre tion data), urban cruise control (using communication with the cisely, a long-term reference velocity is computed based on nfrastructure), cooperative adaptive cruise control (using com- static and slowly changing information (as discussed in See munication between vehicles). We then move to lateral control. tion 5.2): safety in closed loop is guaranteed tracking this refer ence with an aco山8 Predictive cruise control. By predictive cruise control, we in dicate a cruise control tracking a reference velocity that is gen- Urban cruise control. V2I communication can provide look erated using preview information 118119: information can ahead information about traffic and signalized intersections in be static(like road grade and speed limits)or dynumic, but the downstream road. The strategies to explore this informa- slowly changing (like trafic speed). As such, the reference tra- tion are well addressed, especially in arterial scenarios where jectory generation is often cloud-aided(i. e. it exploits informa- the vehicle is driving in traffic through a series of traffic lights tion that is generally retrieved from the cloud) and can be cast When the vehicle receives signal phase and timing information, as an optimization problem. A closely related problem is eco- MPC strategies for ACC have shown substantial energy saving driving, which is concerned with velocity trajectory optimiza- [118129 116]. Compared to a standard ACC, the signal in- tion for minimum energy consumption; this aspect is discussed formation introduces additional position-dependent constraints in Section 5.2 This reference trajectory is based on long-term to enforce that the downstream intersection is crossed during a forecasts and cannot be implemented in open loop. The real- green phase. Generally, the MPC implemented on-board has a time control simply tracks the reference signal, does not exploit limited prediction horizon both because the v2i communica perception sensors or cooperation, and requires driver interven- tion range is limited, and to reduce the computational burden tion to ensure basic safety. Nonetheless, reference generation As discussed in Section 5.2 it is possible to use statistical and for predictive cruise control can also be integrated with any of historical signal data to(remotely)compute a reference velocity the advanced cruise controls discussed next with a long horizon. This reference velocity can be tracked by the on-board urban cruise control, which ensures safety using Adaptive cruise control(ACC). ACC is an enhanced cruise the real -time perception and v2I data control, which detects any preceding vehicle and adjusts speed to the enhancement of passenger safety and comfort. and to operative Adaptive Cruise Control (CACC). CACC is an in order to avoid collisions [120 121. acc design is oriented nhancement of ACC enabled by communication. The per broader impacts like improved road throughput and energy ef formance of ACC is limited by perception systems, that(even fic ciency in the absence of noise and delays) can only measure the rel- MPC has proven effective in simultaneously guaranteeing ative distance and velocity. v2V communication enables the CC safety and performance. In MPC, safety in closed loop exchange of vehicle acceleration(and potentially of its fore is closely related to the problem of persistent feasibility[1221 cast), which can be extremely valuable in dynamic driving sce which is related to the choice of the terminal cost and con- narios. CACC can exploit this additional piece of informa straints; choosing the terminal set as a control invariant set can tion to guarantee higher safety and smaller inter-vehicular dis ensure stability and persistent feasibility. Computing a control tances[130,131 133]. In addition to improved safety, this invariant set is not trivial in the presence of nonlinear dynam can translate into lower energy consumption [134], higher road ics and time-varying, non-convex state constraints. In ACO throughput [130, and passenger comfort: in [135], passengers a conservative approximation is to assume that the preceding using a CACC were found to be comfortable with inter-vehicle vehicle can fully brake at any time; in practice, this can turn time gaps between I s and o6s, while with ACC the acceptable out to be too conservative. Also notice that the preceding ve- time gap was between 2s and ls. v2v and cooperative lon- hicle forecast is uncertain; while certainty equivalence can be gitudinal control have applications in any driving scenario, but adopted, robust and stochastic formulations may be more sys most of the literature is focused on highway driving. Experi- tematic. We refer to [123 61 for a more detailed discussion mental demonstrations are described in [121 22,23 24251 on this. An important role is also played by the inter-vehicula For control analysis and synthesis, the multi-vehicle forma spacing policy: most systems adopt a constant distance policy tion can be regarded as a one-dimensional networked dynamic or a constant heading time policy [121 124125], as we dis system. Much research has been devoted to multi-agent consen cuss further in the next paragraph. While guaranteeing safety, sus schemes, regarding CaCC as a distributed control problem t various performance objectives can be pursued, such as road here we just give an overview of the main challenges, and refer 1 throughput([121. fuel economy [124 126 and driver comfort to 36) for a recent and detailed survey on CACC. As pro- by mimicking her driving style[1231 In applications oriented to energy efficiency improvement,a addressed by classifying the Cav platoon problem dependin common approach is to pursue a small inter-vehicle gap; at high on the choice of speed, this can reduce the aerodynamic resistance In open- road Dynamics, i.e. the dynamics of each CAv experiments with a platoon of trucks, fuel reductions up to 7%0 have been registered [127; for the case of compact vehicles, Information flow network, i. e. the topology and quality of a study with one-eighth-scale models showed considerable re information flow, and the type of information exchanged duction of fuel consumption [128 Combinations of ACC and Figure 4 depicts some typical communication topologies ●+0++ 0+·+●+→ for energy efficiency have highlighted an inherent trade off be (a) Predecessor Following (b)Bidirectional tween air drag reduction(via reduced inter vehicular distance) and powertrain efficiency. More precisely, this trade off is likely to be significant when the velocity profile is variable: maintain ing a small gap may require aggressive throttling and braking, (c)Predecessor Following Leader Bidirectional leader and may lead to suboptimal operation of the powertrain. In 1144, this problem is studied for heavy duty vehicles, when the speed variability is due to road grade, the proposed solu (e)Two Predecessor Following. (f) Two Predecessor Following Leader. tion includes a centralized high-level(cloud-based generation of a speed reference, and a decentralized vehicle-level track- Figure 4: Information low topologies in a four vehicle platoon. The red node ing controller; similarly to[ 123 robust invariance is used to indicates the platoon leader. The nomenclature is taken from(1361 ensure closed loop safety in the CACC In [1261451, the prob lem is approached for light duty CAVS, using forecasts of the used in platooning. The information exchanged may just preceding vehicles velocity; different MPC formulations are be the current velocity and acceleration or include fore possible, depending on the availability of a powertrain model. casts thereof and information on lateral motion Other cacc approaches for energy efficiency were presented in471484950)1 Local controller design and its use of on-board informa- CACC is fundamentally a tracking control problem with tion forecast: as such it has been addressed with a variety of con trol techniques [136]. Linear consensus control and distributed Formation geometry, i.e. vehicle ordering, cruising speed, robust control techniques enable insightful theoretical analysis and inter-vehicle distance and can provide guarantees of string stability [15052]; a pit Notice that the dynamics and the distributed controller pertain fall is the limitation of the dynamics to the linear domain, and the lack of guarantees in the presence of constraints. MPC can to the individual cav. while the information fow network and incorporate nonlinear dynamics, input and state constraints, and the formation geometry are properties of the platoon. The lat- forecasts[153144, 145 A distributed MPC formulation ter two can be decided a priori in a specific demonstration, but require some form of standardization for operation on public suited for any information flow topology, has been presented in roads. A possibility is to coordinate remotely the information How network(based on the instrumentation and on the num ber of vehicle involved) and the formation geometry(based on vehicle characteristics, origins and destinations We further Lateral vehicle control. Lateral control supervises the vehicle discuss this point in Section 5.3 motion in the lateral direction, actuating the steering angle or To ensure safety in closed loop, each CAV must be stable torque. Generally, lateral controllers track with sufficient robustness margins. In an MPC setting, safety tory or path from the motion planning block(described in Sec- can be addressed as in ACC[1391, although with reduced con- tion 4.p, ensuring safety and robustness to model uncertainty servatism thanks to v2v communication. However distur- and a fast changing environment bances acting on the platoon leader may still be amplified in MPC has been fruitfully employed for lateral control, due the downstream vehicles; this phenomenon is known as string to its ability to handle constraints and complex vehicle dynam instability [140 141]. String stability is a property of the local ics; for example, a nonlinear bicycle model was used in[) distributed controller, but it has been shown to depend on the [157] presents an MPC for integrated longitudinal and lateral information flow network and the formation geometry [141]. A control using a linear time-varying model. The MPc-based crucial aspect is the inter-vehicle spacing policy, i. e the choice lateral control in [158] uses a linearized conservative lateral of a reference relative distance d* between two consecutive dynamics model and a overreacting lateral dynamics model to CAVs[,143). Most works in the literature adopt a simple account for two extreme cases in lateral cornering. In 1591 constant distance policy d*= d' or a constant time headway a piece-wise affine model is used for trajectory stabilization in policy the active steering system Some works are specifically focused on the lateral control v十 of CAVs. The lateral controllers in [160, 161 track the lateral where ti is the constant time headway from the preceding ve- motion of the platoon leader. In [162), an MPC-based lateral hicle, v is the current speed ego CAV, and d is a constant min- controller uses vehicular communication to enhance satety in imum distance motion planning and control Looking at the overall system performance and broader im Lateral control design is deeply intertwined with motion pact of platooning, the inter-vehicular spacing or heading time planning; in fact, both algorithms are often based on the same is generally regarded as the main metric. A minimum gap max- models and measurements, Forecasts from communication af- imizes the road throughput [1341331441451 and can re- fect lateral motion also through the motion planning block, as duce the vehicle air drag [130 146]. Recent studies on CACc discussed in Section 4.3 【实例截图】
【核心代码】