Collision avoidance for general aviation

Abstract

This presented the development and simulation of optimized collision avoidance logic to operate on-board general aviation aircraft. Framing the collision avoidance problem as a Markov decision process allows the logic to be generated using a specified encounter model and performance metrics. Past work has focused on using this framework to develop an improved collision avoidance system for non-GA aircraft. This presented results using the framework to optimize logic for lower performance GA aircraft and comparing its performance to the existing version of TCAS, as well as a simple Descend/Climb responsive logic. The results presented herein show that the optimized logic, issuing only climb, descend, and level-off advisories, was able to outperform the current version of TCAS and the Descend/Climb responsive logic against intruders that are equipped with TCAS. The optimized logic issuing 1000 feet = min advisories resulted in the best performance with respect to safety and alerting. For encounters where two GA aircraft are equipped with a collision avoidance system, the optimized logic was 3.1 times safer than TCAS for 500 feet = min advisories and 3.6 times safer for 1000 feet = min advisories. Both of the optimized methods resulted in lower probabilities of alert and reversal than TCAS and the Descend/Climb logic, which is desirable from an operational standpoint. The optimized logic that resulted in the best performance had a relatively high cost of reversing. This resulted in a very low probability of reversing for the GA aircraft, as low as zero in one case. The reversal rate for a TCAS-equipped intruder aircraft was slightly higher, but this was offset by a lower probability of strengthening advisories on the TCAS aircraft.