Control of Multi-Vehicle Systems: Safe, Rapid On-Deck Traffic and Aerial Pursuit
Abstract
The management of the airspace around a ship and the management of traffic of aircraft on board of an aircraft carrier are very similar control problems. They combine two often mutually opposed objectives: rapid arrival at a destination/target and avoidance of collision with other friendly vehicles. Such highly constrained nonlinear optimization problems defy solution in real time, especiallyunder considerable uncertainty coming from the environment and other factors arising in combat. Several recent developments in control algorithm design, pioneered by the PI, offer the promise of a collision-free and rapid guidance to destinations/targets. For achieving rapid reaching of desired destinations in on-deck traffic, or of flying targets, the PI#s new technique of #prescribed-time# stabilization, using time-varying feedback, provides feedback controllers with guarantees of convergence in arbitrary user-prescribed time. These controllers are rigorous generalizations of the classical Proportional Navigation feedback from missile guidance. Theyoutperform sliding mode control and other nonsmooth techniques by making the convergence time independent of the initial condition (i.e., of the vehicle position and orientation), rejecting disturbances of unknown bounds (both deterministic and stochastic), and ensuring jerk-free arrival at the target, when desired (as in on-deck aircraft traffic). For guaranteeing a collision-free operationof vehicles and teams, namely, #safety,# the PI#s recently introduced techniques of prescribed-time safety, fixed-time safety, and inverse optimal safety greatly expand the control designer#s toolkit for high-performance guidance of ground and air vehicles. Existing techniques for ensuring #asymptotic safety# or #exponential safety,# using control barrier functions (CBF) and quadratic programming (QP) suffer from several limitations: (1) needless sacrifice of performance, (2) lack of optimality over the time interval of operation, (3) lack of design flexibility. These limitations are recognized in the literature by the creators of these methods themselves. The PI addresses the collision-free design of on-deck traffic controllers and aerial multi-vehicle engagement scenarios using his new methodologies for the design of prescribed-time and fixed-time #safety filters.# These control algorithms act as overrides on the user#s nominal control actions, as is the case with QP-CBF controllers, but much less conservatively. Rather than keeping the system from ever getting close to the obstacles (i.e., in finite time), these prescribed- and fixed-time safety filters permit the operator to get close to the obstacle in finite time, so as to accomplish the primary task of reaching the desired destination/targetas fast as possible. As in high-performance driving/racing. Additionally, the PI#s techniques of #inverse optimal# safety filter design ensure that, in addition to guaranteeing safety, the control input is not only the closest possible to the user#s commanded input at each time instant, without anticipating the deviation in the future, but guarantee that, over the entire period of operation, the user#s command is followed as closely as safely feasible. This inverse optimal approach enhances both the performance and safety. Furthermore, the inverse optimal approach does not produce only a single, inflexible control formula. It provides a family of safety filters with a user-tunable tradeoff between safety and performance. The project will design controllers for on-deck traffic management and for n-on-m aerial target pursuit using a variety of models: fully actuated, nonholonomic, dynamic with uncertain aerodynamics, in 2D and 3D, with obstacles that move and whose #safety bubbles# account for not only the position but also for orientation and the state of the propulsion system. The control algorithms will be tested with mobile robots in the PI#s lab and in a 2,500 cubicmeter Aerodrome at the PI#s
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 12, 2023
- Source ID
- N000142312376
Entities
People
- Miroslav Krstić
Organizations
- Office of Naval Research
- United States Navy
- University of California, San Diego