Alessandro Colombo

Efficient algorithms for collision avoidance at intersections

We consider the problem of synthesising the least restrictive controller for collision avoidance of multiple vehicles at an intersection. The largest set of states for which there exists a control that avoids collisions is known as the maximal controlled invariant set. Exploiting results from the scheduling literature we prove that, for a general model of vehicle dynamics at an intersection, the problem of checking membership in the maximal controlled invariant set is NP-hard. We then describe an algorithm that solves this problem approximately and with provable error bounds. The approximate solution is used to design a supervisor for collision avoidance whose complexity scales polynomially with the number of vehicles. The supervisor is based on a hybrid algorithm that employs a dynamic model of the vehicles and periodically solves a scheduling problem.

The Two-Fold Singularity of Discontinuous Vector Fields ∗

When a vector field in

Least Restrictive Supervisors for Intersection Collision Avoidance: A Scheduling Approach

\mathbb{R}^3

Nondeterministic Chaos, and the Two-fold Singularity in Piecewise Smooth Flows ∗

is discontinuous on a smooth codimension one surface, it may be simultaneously tangent to both sides of the surface at generic isolated points (singularities). For a piecewise-smooth dynamical system governed by the vector field, we show that the local dynamics depends on a single quantity—the jump in direction of the vector field through the singularity. This quantity controls a bifurcation, in which the initially repelling singularity becomes the apex of a pair of parabolic invariant surfaces. The surfaces are smooth except where they intersect the discontinuity surface, and they divide local space into regions of attraction to, and repulsion from, the singularity.

Teixeira singularities in 3D switched feedback control systems

We consider a cooperative conflict resolution problem that finds application, for example, in vehicle intersection crossing. We seek to determine minimally restrictive supervisors, which allow agents to choose all possible control actions that keep the system safe, that is, conflict free. This is achieved by determining the maximal controlled invariant set, and then by determining control actions that keep the system state inside this set. By exploiting the natural monotonicity of the agents dynamics along their paths, we translate this problem into an equivalent scheduling problem. This allows us to leverage existing results in the scheduling literature to obtain both exact and approximate solutions. The approximate algorithms have polynomial complexity and can handle large problems with guaranteed approximation bounds. We illustrate the application of the proposed algorithms through simulations in which vehicles crossing an intersection are overridden by the supervisor only when necessary to maintain safety.

Discontinuity Induced Bifurcations of Nonhyperbolic Cycles in Nonsmooth Systems

A vector field is piecewise smooth if its value jumps across a hypersurface, and a two-fold singularity is a point where the flow is tangent to the hypersurface from both sides. Two-folds are generic in piecewise smooth systems of three or more dimensions. We derive the local dynamics of all possible two-folds in three dimensions, including nonlinear effects around certain bifurcations, finding that they admit a flow exhibiting chaotic but nondeterministic dynamics. In cases where the flow passes through the two-fold, upon reaching the singularity it is unique in neither forward nor backward time, meaning the causal link between inward and outward dynamics is severed. In one scenario this occurs recurrently. The resulting flow makes repeated, but nonperiodic, excursions from the singularity, whose path and amplitude is not determined by previous excursions. We show that this behavior is robust and has many of the properties associated with chaos. Local geometry reveals that the chaotic behavior can be eliminated by varying a single parameter: the angular jump of the vector field across the two-fold.

Robust multi-agent collision avoidance through scheduling

This paper is concerned with the analysis of a singularity that can occur in three-dimensional discontinuous feedback control systems. The singularity is the two-fold — a tangency of orbits to both sides of a switching manifold. Particular attention is placed on the Teixeira singularity, which previous literature suggests as a mechanism for dynamical transitions in this class of systems. We show that such a singularity cannot occur in classical single-input single-output systems in the Lur’e form. It is, however, a potentially dangerous phenomenon in multiple-input multiple-output switched control systems. The theoretical derivation is illustrated by means of a representative example.

Supervisory control for intersection collision avoidance in the presence of uncontrolled vehicles

We analyze three codimension-two bifurcations occurring in nonsmooth systems, when a nonhyperbolic cycle (fold, flip, and Neimark–Sacker cases, in both continuous and discrete time) interacts with one of the discontinuity boundaries characterizing the systems dynamics. Rather than aiming at a complete unfolding of the three cases, which would require specific assumptions on both the class of nonsmooth system and the geometry of the involved boundary, we concentrate on the geometric features that are common to all scenarios. We show that, at a generic intersection between the smooth and discontinuity induced bifurcation curves, a third curve generically emanates tangentially to the former. This is the discontinuity induced bifurcation curve of the secondary invariant set (the other cycle, the double-period cycle, or the torus, respectively) involved in the smooth bifurcation. The result can be explained intuitively, but its validity is proved here rigorously under very general conditions. Three examples f...

Supervisory control of differentially flat systems based on abstraction

We propose a class of computationally efficient algorithms for conflict resolution in the presence of modeling and measurement uncertainties. Specifically, we address a scenario where a number of agents, whose dynamics are possibly nonlinear, must cross an intersection avoiding collisions. We obtain an exact solution and an approximate one with quantified error bound whose complexity scales polynomially with the number of agents.

Two degenerate boundary equilibrium bifurcations in planar Filippov systems

This paper describes the design of a supervisory controller (supervisor) that manages controlled vehicles to avoid intersection collisions in the presence of uncontrolled vehicles. Two main problems are addressed: verification of the safety of all vehicles at an intersection, and management of the inputs of controlled vehicles. For the verification problem, we employ an inserted idle-time scheduling approach, where the “inserted idle-time” is a time interval when the intersection is deliberately held idle for uncontrolled vehicles to safely cross the intersection. For the management problem, we design a supervisor that is least restrictive in the sense that it overrides controlled vehicles only when a safety violation becomes imminent. We analyze computational complexity and propose an efficient version of the supervisor with a quantified approximation bound.