Vehicle platooning pdf

Create Alert Alert. Share This Paper. Nonlinear spacing policy based vehicle platoon control for local string stability and global traffic flow stability. Engineering, Computer Science. Computer Science, Engineering. Experimental vehicles: two Cycab leading two RobuCab earization techniques offer a relevant framework to address platoon control: equations 1 , as most of kinematic models In this paper, the kinematic tricycle model is considered: of mobile robots, can be converted in an exact way into a so- the two actual front wheels are replaced by a unique virtual called chained form, see [18].

Such a conversion is attractive, wheel located at the mid-distance between the actual wheels. Details can be found in [19]. In nominal situation, the objective for tion are called control points. Once more, exact linearization techniques data available at instant T.

However, in order to limit Cycab and RobuCab vehicles see Fig. Two additional advantages are a bounded computing time III. More precisely, let nac as active curves be the number si. As a result, the reference trajectory representation must of polynomials Qi t entering into the optimization process. Finally, let nap as active points be the number of control In order to meet these requirements, it is here proposed to points Pxk , Pyk whose coordinates are to be modified rely on B-Spline curves.

The raw data from which these when minimizing 4. Approximation of raw data by B-Spline is first Therefore, in order to freely shape nac polynomials Qi t recalled, and then the extension process is described. Each polynomial Qi t is a linear combination damaged, and consequently the overall fitting performances.

The performances and thus the overall one can also be damaged. Implementation details line reference trajectory generation to its main parameters, When a new localization data is available, two possibilities and eventually propose their most suitable values.

Then, full- have to be considered. They are illustrated in Fig. A new control point is added, when the 20 10 oldest active one is fixed, see bottom-figure 3. Finally, in order to be able to compute the arc-length coor- Fig. Trajectory for simulations Fig. Trajectory for experiments dinate s1 of the first vehicle between two raw acquisitions, First, a posteriori optimizations i. Clearly, parameter c2 has before minimizing criterion 4 , see Fig.

Since raw localization data have here been recorded at a 10Hz sampling frequency, with a limited measurement noise the accuracy of the RTK GPS is 2cm , when the vehicle was driven at a velocity of 1m.

The average and the maximum euclidian distances between the raw data i. As it could have been intuitively expected, whatever the value of d is, the best approximations are obtained with the smallest value of c2. Moreover, they are very satisfactory, since the absolute average error is less than 5mm. B-Spline extension process 4 3 0. Influence of nac and nap on on-line optimization Next, on-line optimizations i.

They are electric vehicles, powered by choices for d, c2 shown in bold in Fig. Two resp. The four passengers can travel aboard the Cycab resp. Their small dimensions length 1. Inter-vehicle communication d , increasing nac value does no longer improve significantly is ensured via WiFi technology. Platoon control laws are the quality of the approximation. This means that modifying polynomials 2 Experimental results: The experiments have been car- Qi t which are not explicitly considered in local crite- ried out with four vehicles.

Consequently, first vehicle and their projection on the B-Spline are resp. These figures are quite similar to what perfectly constant. However, the average absolute variations was expected from simulation trials, see Fig. And once more, the best results 0. However, as can be seen in Fig. Experiments Fig. Error in the approximation of the first vehicle trajectory Several experiments have been carried out in Clermont- The lateral deviations of each vehicle w.

It can be observed urban transportation system evaluation. The video attachment that the behavior of vehicle 2 is slightly different from those presents some sequences recorded during the experiments. Then, nonlinear control tech- vehicle 2 since it is close to the leader , when it has become niques have been considered to take explicitly into account constant for vehicles 3 and 4.

However, the lateral guidance the nonlinearities in vehicle models, in order to enable high of vehicle 2 is roughly as accurate as for vehicles 3 and 4, and accurate guidance. Finally, full scale experiments, carried out as satisfactory as in previous work, when all vehicles were with four vehicles, have demonstrated the efficiency of the guided w.

Current developments aim at refining the on-line refer- ence path generation: enhanced criteria, with adaptive subset 0. Benhimane and E. Wang and M. Unified control design for autonomous car-like 60 80 vehicle tracking maneuvers.

Consolini, F. Morbidi, D. Prattichizzo, and M. Leader- Fig. Vehicle lateral deviations follower formation control of nonholonomic mobile robots with input constraints.

In other embodiments, the processing capability required by the system of FIG. In appropriate embodiments, such tablets or smartphones can serve as the core controller, the user interface panel, or can provide some or all of the vehicle-to-vehicle link through either cellular connectivity, Bluetooth, WiFi, or other suitable connection.

The user receives information a from visual and or auditory alerts, and can make system requests e. The user interface communicates with a long range data link b , such as through a cellular modem or other service. The user interface is responsible for managing this data link, sending data via the link, and receiving data via the link. A control processor which may be alternatively integrated with the GUI box receives sensor information c , short range data link information e , and controls the actuators f.

Alternately, the long range communication link can connect directly to the control box In this case, the user interface may be an extremely simple low cost device, or may even be eliminated from the system entirely.

In particular, and with reference to FIG. Via connection a , typically but not necessarily a CAN interface, the control processor configures the radar unit and receives data. Connection b to accelerometers , which can be wireless, gives the control box acceleration information in 1, 2 or 3 axes as well as rotation rate information about 1, 2 or 3 axes.

In some embodiments, gyros can be substituted for accelerometers, especially for, for example, rotation rate. The data link , shown at c and illustrated in greater detail below as indicated by the dashed lines, provides information about relevant characteristics of the leading truck , including its acceleration, or is used to provide the same or similar information to a following truck The brake valve d provides data on brake pressure, and is used to apply pressure via a command from the control processor The accelerator command is sent via an analog voltage or a communications signal CAN or otherwise.

The control processor performs calculations to process the sensor information, information from the GUI, and any other data sources, and determine the correct set of actuator commands to attain the current goal example: maintaining a constant following distance to the preceding vehicle.

The data links can comprise a link to the truck manufacturer's engine control unit , a wireless link for communications and a link to other aspects of the vehicle as shown at Each of these links can, depending upon the embodiment, be bidirectional. The operation of the vehicle control unit of the present invention can be better appreciated from FIG. Two modes of operation are possible: in a first mode, the front truck's control unit issues commands to the back truck's control unit, and those commands are, in general, followed, but can be ignored in appropriate circumstances, such as safety.

In a second mode, the front truck's control unit sends data to the second truck, advising the trailing truck of the data sensed by the lead truck and the actions being taken by the lead truck.

The second truck's control unit then operates on that data from the front truck to take appropriate action. As shown at , the following or trailing truck sends data about its operation to the front or lead truck. The lead truck then decides upon actions for the lead truck, shown at , and, if operating in the first mode, also decides upon actions for the back truck, shown at Then, depending upon whether operating in first or second mode, the lead truck either sends commands to the trailing truck first mode , or sends data to the trailing truck second mode.

If operating in the first mode, the second truck receives the commands and performs them at , with the caveat that the second truck can also chose to ignore such commands in some embodiments. If operating in the second mode, the second truck receives the data at , and decides what actions to perform. Because the control programs for both units are the same, in most cases the resulting control of the second truck will be identical regardless of operating mode.

Finally, the second truck communicates to the front truck what actions it has taken, so that each truck knows the state of the other. In at least some embodiments, this process is repeated substantially continually to ensure that each truck has the current state of the other truck, thus ensuring safe and predictable operation of each truck, even when operating in close-order formation at highway speeds. First, the system identifies a vehicle available for coordination example: within a certain range, depending on the route of the two vehicles.

Once one of the vehicles has accepted or , the other can then accept, meaning that the pair has agreed to coordinate for possible linking Depending on vehicle positioning, weight of load, vehicle equipment, and other factors, a vehicle within linking range may be identified as a Following Vehicle Available for Linking or a Leading Vehicle Available for Linking If neither of these is the case, the system returns to coordination mode.

Once a Following Vehicle Available for Coordination has Accepted the link , the Self Vehicle then also accepts the link , initiating the link. Upon completion of the link, the vehicles are now linked Similarly, once a Leading Vehicle Available for Coordination has Accepted the link , the Self Vehicle then also accepts the link , initiating the link.

In FIG. Because, in at least some embodiments, the vehicle control unit knows a variety of details about the truck on which the system is installed including either torque, engine speed, and acceleration, or power and acceleration the engine torque and acceleration permits vehicle mass to be calculated, shown at Based upon that calculation for each truck in the pair, the trucks are determined either to be suitable for linking, or not. If they are suitable for linking, shown at , a determination as to which truck should lead is made at , using the factors mentioned above.

In some instances, the characteristics of the truck, such as load, etc. Further, if an exit, interchange, or other road feature or condition is encountered, or is being approached for example, as detected by vehicle sensors or communicated from an external source such as the fleet office then the distance can be increased to provide visibility to the rear driver.

Additionally for an upcoming exit the rear truck or both trucks can be set to coast to avoid braking at the off-ramp. In some embodiments, the following distance can also be adjusted based on other upcoming features of the road or greater environment, to ensure safety, make the driver more comfortable, or for other reasons.

Dangerous low overpasses, inspection stations, road grade, or areas identified as dangerous, can all be used to adjust the following distance. These features can be identified from map data, internet data, or other source.

Other features can be detected by either or both trucks, either from their on-board sensors, or from the sensors added for the system. These include upcoming road curvature, current or upcoming road grade. Current or upcoming traffic can also be identified through radar sensors, the internet, machine vision, or other methods. In some embodiments, the following distance can also be set based on driver activity.

A lack of steering input can signify inattention and cause an increase in following distance. Similarly, aggressive behavior, shown by aggressive motion of the steering wheel, pedals or other input, can be used to set a desired distance. The turn signal can also change the distance, for example to allow space between the vehicles for exiting the road.

The driver can also select the following distance in some embodiments. Still further, the current fuel economy, the amount of fuel onboard, the projected range, or other fuel-related parameters may be used to set the following distance. For example the driver may want to follow more closely when the fuel level is low, to help reach a destination.

As another example, the fleet or the driver may have a target fuel economy, and the adjustment of following distance can be used to meet this target, within limits appropriate to ensuring safety. In the event the leading vehicle is required to make emergency maneuvers, safety is ensured by the use of the communications link between the two vehicles. The data link has very low latency approximately 10 ms in one embodiment , and high reliability.

This link could also be a non-industry-standard format. In the event of a data link loss, the trailing vehicles are typically instructed to immediately start slowing, to ensure that if the front vehicle happens to brake immediately when the link is lost, the gap can be maintained safely. In addition to safe operation during the loss of the data link , the system should be safe in the event of failure of components of the system.

For most failures, the trailing vehicles start braking, until the driver takes control or other sensors determine that the situation is safe at which point braking can be decreased as appropriate.

This ensures that, in the worst case where the front vehicle starts to brake immediately when a system component fails, the system is still safe. The modified brake valve is also designed such that in the event of a complete failure, the driver can still brake the vehicle. Ordering of the vehicles: In an embodiment, the system arranges the vehicles on the road to ensure safety. In such an embodiment, the system will graphically or otherwise tell the drivers which vehicle should be in the front.

For example, to mitigate fatigue, the system may cause the trucks to exchange positions on a periodic basis. In embodiments where order is important, such as heavy trucks, the system will only perform the linking functionality if the vehicles are in the correct order. The order may be determined by relative positioning measures like GPS, directional detection of the wireless communication, driver input, visual video or still image processing, or direct or indirect detection of aerodynamics through fuel savings or sensors.

In another embodiment, the system can apply steering or other lateral control, combined with control of engine torque and braking, if needed, to effectuate the desired order of the vehicles. One such feature is to use the video stream from the front looking camera to detect drifting within or out of the lane. This is done by looking at the edges or important features on the leading vehicle , and calculating the lateral offset from that vehicle.

When it is detected, the system can react with a braking jerk a short braking application to get the driver's attention , slowing down, or a braking jerk in the leading vehicle. Alternatively, and as shown in FIG. In embodiments having video, portions of the video that are not important, or change less frequently, can be highly compressed or not transmitted at all. For example, when trucks are linked, the back of the lead vehicle does not change significantly, and is not critical.

The compression can be varied based on known or commanded movement of the vehicles. For example if it is known that the vehicles have relative motion laterally, then the image can be shifted laterally in an efficient way without sending the raw video. The system can also use the front mounted radar to detect obstacles or stationary vehicles in the road, even when not in close-following mode.

When these are detected, it can apply a braking jerk, slow the vehicle, or provide visual or auditory warnings. The system can also use the accelerator pedal signal to determine when the driver is not engaged with the vehicle or other driver states and react accordingly, such as slowing the vehicle or disabling the system.

These and other warnings and alerts are discussed hereinafter in connection with FIG. To facilitate rapid deployment, a simpler version of the system enables vehicles to be a leading vehicle, shown in FIG. The components on this version are a subset of those on the full system, so there is no automation. There are several embodiments of this reduced set of functionality, with different subsets of the components from the full system.

As such, this version has only the data link functionality It connects to the brake pressure sensor and electrical power. The full system may also provide other fuel economy optimizations. These may include grade-based cruise control, where the speed set-point is determined in part by the grade angle of the road and the upcoming road. The system can also set the speed of the vehicles to attain a specific fuel economy, given constraints on arrival time. Displaying the optimum transmission gear for the driver can also provide fuel economy benefits.

The system may also suggest an optimal lateral positioning of the trucks, to increase the fuel savings. For example, with a cross wind, it may be preferable to have a slight offset between the trucks, such that the trailing truck is not aligned perfectly behind the leading truck. This lateral position may be some combination of a relative position to the surrounding truck s or other vehicles, position within the lane, and global position.

This lateral position may be indicated by the registration marks The data link between the two vehicles is critical to safety, so the safety critical data on this link has priority over any other data.

Thus the link can be separated into a safety layer top priority and a convenience layer lower priority. The critical priority data is that which is used to actively control the trailing vehicle. For example in an emergency braking situation, additional data may be included as high priority. The lower priority convenience portion of the link can be used to provide data, voice or video to the drivers to increase their pleasure of driving.

This can include social interaction with the other drivers, or video from the front vehicle's camera to provide a view of the road ahead. This link can also be used when the vehicle is stationary to output diagnostic information gathered while the vehicle was driving. In addition, other cameras, and thus other views, can be provided, including providing the driver of the lead truck with a view from the forward-looking camera on the trailing rig, or providing both drivers with sufficient camera views from around each vehicle that all blind spots are eliminated for each driver.

Because the system is tracking the movements of the vehicles, a tremendous amount of data about the individual vehicles and about the fleet is available. This information can be processed to provide analysis of fleet logistics, individual driver performance, vehicle performance or fuel economy, backhaul opportunities, or others. These and other features are discussed hereinafter in connection with FIGS.

The button triggers an increase in the vehicle gap to a normal following distance, followed by an automatic resumption of the close following distance once the merging vehicle has left. The length of this gap may be determined by the speed of the vehicles, the current gap, an identification of the vehicle that wishes to merge, the road type, and other factors.

The transition to and from this gap may have a smooth shape similar to that used for the original linking event. For vehicles without an automatic transmission, the system can sense the application of the clutch pedal by inferring such from the engine speed and vehicle speed. If the ratio is not close to one of the transmission ratios of the vehicle, then the clutch pedal is applied or the vehicle is in neutral. In this event the system should be disengaged, because the system no longer has the ability to control torque to the drive wheels.

Thus if none of these are true, the clutch pedal is engaged. The system can estimate the mass of the vehicle to take into account changes in load from cargo. The system uses the engine torque and measured acceleration to estimate the mass. This may also be combined with various smoothing algorithms to reject noise, including Kalman filtering, Luenberger observers, and others. This estimate is then used in the control of the vehicle for the trajectory generation, system fail-safes, the tracking controller, and to decide when full braking power is needed.

The mass is also used to help determine the order of the vehicles on the road. Many modifications and additions to the embodiments described above are possible and are within the scope of the present invention. For example, the system may also include the capability to have passenger cars or light trucks following heavy trucks. This capability may be built in at the factory to the passenger cars and light trucks, or could be a subset of the components and functionality described here, e.

The system may also include an aerodynamic design optimized for the purpose of convoying, as shown in FIG. This may be the design of the tractor or trailer, or the design of add-on aerodynamic aids that optimize the airflow for the convoy mode.