Abstract
Hybrid systems are characterized by having an interaction between continuous dynamics and discrete events. The contribution of this paper is to provide hybrid systems with a novel geometric formulation so that controls can be added. Using this framework we describe some new global controllability tests for hybrid control systems exploiting the geometry and the topology of the set of jump points, where the instantaneous change of dynamics take place. Controllability is understood as the existence of a feasible trajectory for the system joining any two given points. As a result we describe examples where none of the continuous control systems are controllable, but the associated hybrid system is controllable because of the characteristics of the jump set.
Introduction
In control theory, controllability determines whether a system can be steered from any initial state to any desired target state using available inputs. While well-understood for purely continuous systems, hybrid systems—which combine continuous motion with discrete "jumps" or transitions—present significant analytical challenges. Traditional approaches often rely on restrictive assumptions about when and where transitions can occur. This paper introduces a rigorous geometric framework to analyze hybrid systems, treating discrete jumps not as obstacles, but as potential resources for achieving global controllability.
Geometric Framework for Hybrid Systems
The authors model hybrid systems using the language of differential geometry. The fundamental building blocks are Generalized Dynamical Systems (GDS), defined as fiber bundles equipped with differentiable maps and submanifolds representing constraints. A complete hybrid system is then represented as a directed graph.
In this model, each node corresponds to a specific continuous dynamic mode, and edges represent transitions between these modes. These transitions are governed by a jump map ($R$), which dictates how the system state is transformed during a discrete event. The formal mapping between these components is defined through commutative diagrams that link the state space, control bundles, and transition manifolds.
Controllability Analysis
The core of the paper's contribution lies in developing tests for global controllability. The authors utilize the Orbit Theorem and the theory of foliations to describe the "leaves" or reachable sets of the continuous dynamics.
The interaction between these continuous leaves and the discrete jump sets is the key to controllability. By applying transversality conditions, the authors show that if the tangent space of the jump set and the tangent space of the dynamical leaves together span the entire configuration space, the system can move in directions that neither the jump nor the continuous motion could achieve in isolation. This allows the system to "hop" between otherwise disconnected reachable regions.
Key Results
The paper presents two primary theoretical results:
- Theorem 4.6: Provides an algorithmic condition for global controllability based on sequences of discrete modes and jump points.
- Theorem 4.13: Offers a practical test stating that if a hybrid system has a strongly connected graph and its continuous leaves intersect jump sets such that their combined tangent spaces cover the manifold, the system is globally controllable.
A notable application is the "Submarine" example. In this scenario, a system is restricted to horizontal movement at specific depths (e.g., due to environmental constraints). While the individual horizontal modes are not controllable in 2D space, the addition of vertical jumps at specific locations makes the entire space reachable, demonstrating how discrete transitions can "gain" controllability for a system.
Conclusion and Future Work
This geometric approach provides a foundation for analyzing complex hybrid systems. However, challenges remain, such as the "dead end" problem where jumps lead to states that violate future constraints. Future research will focus on hybrid time discretization for computer simulations and extending this framework to address optimal control—finding not just any path, but the most efficient one.
Summary updated