| dc.description.abstract | Adaptive cruise control (ACC) testing requires minimum of two cars and a platform where the two cars can be tested for a continuous time. Here a custom-built platform and software are presented for testing various ACC algorithms on scaled model cars. There are multiple techniques being studied for driver convenience and safety automation systems for production vehicles: electronic stability control, adaptive cruise control, lane keeping, and obstacle avoidance. Presented here are some novel control framework that gives formal guarantees of correctness that go beyond traditional PID-based controllers for ACC that do not, inherently, have proofs that satisfy. In the first approach, safety constraints – maintaining a valid following distance from a lead car are represented by control barrier functions (CBFs), and control objectives – achieve a desired speed – are encoded through control Lyapunov functions (CLFs). While the same safety constraints are formulated using Linear Temporal Logic (LTL) for synthesizing the control software module using abstraction based controllers in the second approach. In the longer run, each interacting software module is endowed with specifications, under certain environment assumptions, the module is guaranteed to meet its specifications.
For the CBF-CLF approach, the different objectives can be unified through a quadratic program (QP), with constraints dictated by CBFs and CLFs that balance safety and control objectives in an optimal fashion. Similarly for the abstraction controllers, PESSOA and Polyhedral Control Invariant Set approaches are correct-by-construction. The end result was the experimental demonstration of these methodologies on scale-model cars, for which the CBF-CLF and abstraction based controllers were implemented in real-time. | en |
