Organization and Control of Autonomous Railway Convoys

More documents

Recommendations

Info

AVEC ’08broadcasting to all involved vehicles. All other vehiclesonly provide data to the direct following vehicle (seeFig. 6). The cycle time T for the radio communication is40 ms.The effects of the wireless communication networklike varying latencies in data transmission and packetlosses cannot be neglected. Additionally an algorithmfor generating data in the step time of the controller hasto be implemented. The principle is shown in Fig. 7.controller. With the gain K p of position controller thefollowing equation has to be fulfilled**∆ x n+ 1( Tswitch) ⋅ Kp= vn+1(Tswitch)(7)In order to avoid discontinuities the distance will bereduced by a ramp signal.A feed forward control improves the dynamicbehavior of the following vehicles. In case of convoymode, the current speed of the convoy leader, which issent to all followers by broadcast communication, isapplied in the feed forward line of all followers and isonly activated in position control mode (see Fig. 8).Fig. 7bufferPrinciple of network controlled system with predictor andA predictor is implemented in addition to thecontroller of the leading vehicle which includes aprecise model of the vehicle dynamics. The predictorestimates the progress of the speed trajectory V ˆ n for adefined prediction horizon based on the track topology.The prospective development of the speed is thenpartitioned into the time pattern of the datacommunicationV ˆn = { vˆn ( T )..ˆ vn( kT )}(6)The predicted vector of the speed is sent to thefollower. The following vehicle is then able to estimateprecisely the behavior of the leader in case of packetlosses or high latencies.The follower is expanded with a Kalman-Filter anda buffer. The filter is used for estimating the position ˆx n ,respectively the distance ∆x ˆn,n+1 , and the speed ˆv n . Thefilter also includes a buffer for the received speedtrajectory Vˆ n . Even if packet losses occur, the filtergets its input every period T within the predictionhorizon of kT. The buffer is overwritten with eachaccurately received data. Within a prediction horizon of10 transmission cycles (400 ms) the accuracy of thedetermined position can be improved up to 8 cm.4.2 Reference GeneratorThe reference generator of a follower RailCab isdifferent than the leader’s one. Additionally, a situationanalysis is executed inside by evaluating the currentposition and the speed of the leader n and the followern+1. An additional input comes from the configurationcontrol which evaluates the system state information.Either a normal state or a hazard occurrence [9] isidentified. The set values for the position control arecalculated by evaluating the braking distances withknowledge of the track topology [7].The reference generator also outputs a signal of theconvoy state which is needed for switching between thecontrollers. However, at the switching instant T switch acontinuous transition is required between speed andposition control. Therefore the output of the positioncontroller has to be equal to the input of the speedFig. 8Structure of convoy controllerFig. 9 illustrates the behavior with the feed forwardcontrol which ensures convoy stability.Fig. 9Proof of convoy stability5. CONVOY CONTROLThe objectives of the convoy control law are a fastformation of a convoy under safety requirements andthe safe operation in platoon mode. A distance controlas well as a speed control is applied for convoy controldepending on the current situation. In this section, thedistance control is described first. Furthermoremaneuver control laws are presented.5.1 Distance ControlThe reference position of the follower is defined as* *x n + 1 = xn− d ( vn, vn+ 1,an, an+ 1)(8)where the distance reference d * depends on the currentoperating mode and includes the length of the vehicle(3.5 m).Some fundamentals of safe vehicle operation aredefined first. Absolute and relative braking distances aredistinguished. The absolute braking distance of thefollower d abs,n+1 is an uncritical distance and defined as2vn+12 n + 1d abs,n+ 1 = (9)a
AVEC ’08This distance also defines the inter-platoon distance. Incontrast, the relative braking distance d rel,n+1 is thedifference if the absolute braking distances of twoRailCabs and is defined asdrel,n 12n+1n+12nnv v+ = −(10)2a2aIt is safety critical to drive within the relativebraking distance. In order to avoid crashes, the vehiclesmust a priori agree upon the maneuvers. However, thiscannot be ensured in all cases. Therefore maneuvercontrol laws have been designed which minimizes riskswhile driving within distances of less than one meter.5.2 Maneuver Control LawsIn general three kinds of maneuvers in case ofconvoy mode can be distinguished: merging andsplitting of convoys and driving in a platoon. Themaneuver control laws are based on the longitudinalcontrol described above. The reference generator adaptsthe reference values depending on the current situationand outputs a convoy state signal for switching betweenthe operating modes as described above.The control strategy is realized as a state modelwith the three main states NOCONVOY,MERGING/SPLITTING and PLATOON as shown in Fig. 10.In order to avoid permanent switching betweenstates, the conditions additional include a margin. Acomplete merging process is shown in Fig. 11. Whereasthe effective distance of 4 m has to be reduced by thevehicle length of 3.5 m.Fig. 11Convoy formation (complete process)PlatoonThe platoon phase is activated by the transitionfrom speed to position control. The criterion is aminimal distance d min which is needed for reducing thespeed difference between the two vehicles. Thus, thedesired distance is smoothly adjusted (see Fig. 12).Fig. 10State machine of the convoy control strategyMergingMerging of vehicles to a platoon is the most safetycritical operation because of unavoidable speeddifferences while driving within small distances.Therefore a control strategy is applied which limits thespeed during this process. Phase 1 of the mergingprocess begins with undershooting the absolute brakingdistance. Here the speed controller limits the speeddifference between follower and leader to 2m/s. Thespeed has to be reduced before the phase begins.The safety critical process starts with reaching therelative braking distance. In phase 2, the speeddifference is limited to 0.7m/s. A maximum speeddifference of 1 m/s when the time delay of activationand engagement of the hydraulic emergency brake(approximately 300 ms) is considered. This speeddifference is assumed to cause no remaining damages incase of a collision.Fig. 12control)Convoy formation (transition from speed to positionIn platoon mode, a fixed distance reference isapplied independent of the current vehicle speed. Thedesired distance is shortly undershot due to a brakingprocedure without propagation by the leader as shownin Fig. 13. This can be avoided with the a prioripropagation of changes of the operational profiles.Fig. 13 Reaction of follower after braking of the convoy leader innormal case
Page 1 and 2: AVEC ’08Organization and Control
Page 3: AVEC ’083. CONTROL OF LONGITUDINA

Organization and Control of Autonomous Railway Convoys

Create successful ePaper yourself

Delete template?

Save as template?