Autopilot Improvements

Map-based Speed Limits: Speed Limit Recognition often fails. Add an option to auto-adjust speed based on map data. Safety: LCC disengagement needs clear audio and visual warning (silent standby mode causes dangerous lane drift) Traffic Jam Assist: Allow longer adjustable interval before traffic jam assist prompts for acceleration Smarter Nags: When the driver monitoring camera sees my eyes on the road, delay the steering wheel nag. (My hand is on the wheel but it can't detect it if there is minimal movement) End-to-End Model: Replace the legacy object detection based architecture with and End-to-End model. This would allow for more natural lateral movements, taking in account the lead car's behaviours and adjacent vehicles. The Vision/Radar system needs attention. Current emphasis on China-exclusive LIDAR hardware neglects globally available systems. While the LIDAR version's Nvidia Orin-X is more powerful, the Snapdragon chip has sufficient performance, OpenPilot 0.8x achieved smooth end-to-end lateral planning back in 2021 on older, weaker hardware. Current limitations (missing curbs, road edges, cones, lead car detection failures causing jerking) can be fixed by switching to end2end architecture. *Autopilot: LCC (Lane Centering L2 ADAS)