- The T265 has stereo fisheye cameras, so in theory you can compute depth from T265 images using OpenCV, then perform obstacle avoidance in some way. In practice, the depth quality will not be reliable enough for obstacle avoidance so you would be better off just use a dedicated sensor appropriate for this task.
- Object avoidance with rotating LiDAR has also been long available, so that would be the go-to solution. There are some 2020 GSoC ArduPilot projects related to obstacle avoidance (using depth camera as well). You can follow and even become a beta tester to boost the progress yourself.