Online Verification of Automated Road Vehicles Using Reachability Analysis
read more
Citations
Planning and Decision-Making for Autonomous Vehicles
A Review of Motion Planning for Highway Autonomous Driving
Dynamics and Control
Funnel libraries for real-time robust feedback motion planning:
Real-Time Trajectory Planning for Autonomous Urban Driving: Framework, Algorithms, and Verifications
References
A time-dependent Hamilton-Jacobi formulation of reachable sets for continuous dynamic games
SpaceEx: scalable verification of hybrid systems
Real-Time Motion Planning With Applications to Autonomous Urban Driving
Monotone control systems
Related Papers (5)
Frequently Asked Questions (16)
Q2. What have the authors stated for future works in "Online verification of automated road vehicles using reachability analysis" ?
In the future, the authors plan to propose a general-purpose model for the set-based prediction of other traffic participants considering a wider range of traffic rules.
Q3. What is the main argument for choosing a fixed time step?
Another advantage of fixed step size is that the occupancies can be more easily synchronized with other traffic participants when a common time step is used, which is the main argument for choosing a fixed time step.
Q4. What are the measurements of the disturbance set Y?
The disturbance set Y is chosen as 0 for all dimensions, except for the dimensions adding uncertainty to β̇ and Ψ̈, which are altered when the tire contact forces vary due to damaged tarmac.
Q5. What is the alternative to calculating the reachable set of the ego vehicle?
An alternative to computing the reachable set of the ego vehicle based on the vehicle dynamics (under consideration of a set of initial states, input trajectories, and a set of parameters), is to simply add a fixed deviation from the reference trajectory.
Q6. What is the simplest way to compute the linearization errors?
The set of linearization errors L in (4) requires the set of reachable states R(τk), which in turn requires the set of linearization errors to be computed.
Q7. How long does the time increment for the scenario in Sec. VI-A vary?
The time increment for the scenario considered in Sec. VI-A (including the attached braking maneuver) varies from 0.0071 to 0.0189 seconds.
Q8. What is the chopping for the left border?
The chopping for the left border is denoted by choplat(OCcompl, s̃l(tk),Hl), where s̃l(tk) is the orthogonal distances to the halfspace Hl = {x|nTl x ≤ dl}, where nl is the normal and dl the distance to the origin.
Q9. What is the behavior of inputs such as noise and disturbances?
Note that the time-varying behavior of inputs such as sensor noise and disturbances is arbitrary, as long as the values are within bounded sets.
Q10. How do the authors prevent inevitable collision states?
The authors prevent inevitable collision states by only accepting intended plans with a subsequent maneuver that brings the vehicle to a stop at a safe location, such that it cannot cause a collision for all future times, see [40, Sec. IV.E].
Q11. How is the used trajectory planner adapted?
The used trajectory planner should be adapted such that new reference trajectories branch off previous ones at points x(tver) that are reached by the ego vehicle when the verification of the new reference trajectory is completed, as illustrated in Fig.
Q12. What are the constraints that are removed when a pedestrian crosses a street?
E.g. when a pedestrian crosses a street where no crosswalk is present, constraint C5 is removed and only constraints C1 and C2 are active.
Q13. Why is the symbol for set-based multiplication omitted?
Note that the symbol for set-based multiplication is often omitted for simplicity of notation, and that one or both operands can be singletons.
Q14. what is the reachable set for the next point in time?
The reachable set for the next point in time and time interval is obtained by combining all previous results and using the operator co(·) for the convex hull:R(tk+1) =eArR(tk)⊕ Γ(r)ûc ⊕Rp(r), R(τk) =co ( R(tk), eArR(tk)⊕ Γ(r)ûc ) ⊕Rǫ ⊕Rp(r)(9)The reachable set makes it possible to compute the set of positions OC(τk) occupied by the vehicle on the road for each time interval τk.
Q15. What is the way to compute the occupancy of other traffic participants?
the dynamics of the model for other traffic participants is monotone under certain conditions and the occupancy can be exactly computed by constraining only the absolute acceleration.
Q16. What is the effect of a deviation of a few centimeters on the vehicle?
Although uncertainties in the movement of the ego vehicle are considerably smaller than the ones of other traffic participants, neglecting uncertainties in the movement of the ego vehicle could cause the vehicle to lose track of the reference trajectory or hit the road boundary, for which a deviation of a few centimeters can be crucial in some situations.