Core Concepts
This paper proposes a risk-aware model predictive control scheme for linear stochastic systems that can dynamically evaluate and accept or reject runtime signal temporal logic specifications while guaranteeing satisfaction of previously accepted specifications.
Abstract
The paper addresses the problem of controlling stochastic systems with temporal logic specifications that can be dynamically assigned during runtime. Conventional risk-aware control typically assumes that all specifications are predefined and remain unchanged during runtime.
The key highlights and insights are:
The authors propose a novel, provably correct model predictive control scheme for linear systems with additive unbounded stochastic disturbances. The control method dynamically evaluates the feasibility of runtime signal temporal logic specifications and automatically reschedules the control inputs accordingly.
The control method guarantees the probabilistic satisfaction of newly accepted specifications without sacrificing the satisfaction of the previously accepted ones. It has the flexibility to reject new specifications if necessary.
The control scheme utilizes probabilistic reachable tubes to relate each temporal logic specification to its maximal risk via (mostly) linear constraints. This allows the tube-based MPC problem to be formulated and solved at each time step.
The authors prove the recursive feasibility of the overall control scheme and show that the open-loop implementation satisfies the probabilistic constraints on the temporal logic specifications.
The proposed control method is validated through a robotic motion planning case study, where the robot dynamically receives and handles new reach objectives during runtime.