A Vision-based Solution to Estimating Time to Closest Point of Approach for Sense and Avoid

Autonomous Agent for Beyond Visual Range Air Combat: A Deep Reinforcement Learning Approach

“…For specific calculation details, see Algorithm 1. The second comfort metric assesses the safety of autonomous aircraft operations by calculating the risk of collisions based on the principles of the Time to Closest Point of Approach (TCPA) [10]. The collision risk comfort metric, 𝑀 5 , increments in situations where the distance 𝑑 between an autonomous aircraft and a wingman is less than a critical threshold 𝜀, recommended to be set at 0.5 nautical miles for initial trials [8], or when the closing velocity 𝑣 close,𝑡 indicates a decreasing distance that could lead to a collision within a critical time frame 𝑡 critical .…”

“…Autonomous aircraft in shared airspace must navigate safely and efficiently while adhering to social norms expected in humancentric environments [9]. These norms include respecting personal space [1], maintaining comfortable velocities and accelerations [11], and keeping a safe distance from other aircraft [10]. Research into socially aware navigation aims to improve interactions between autonomous agents and humans [12], but there is still a need for new metrics to evaluate these methods more effectively [14].…”