Idea
Framework improving autonomous vehicle safety and trajectory planning by filtering unreliable sensor data and enhancing scene understanding.
Research Paper
Core Innovation
This paper introduces RCT-AD, which uniquely integrates a reliability scoring module with a quality-gated memory mechanism to filter and reconstruct degraded sensor inputs. It combines this with a temporal trajectory planner and joint detection-segmentation to improve scene representation and planning robustness over prior BEV approaches that fuse temporal data indiscriminately.
Why It Matters
Autonomous vehicles face frequent sensor degradation in urban environments, risking unstable planning and collisions. This framework enhances reliability by filtering corrupted data and leveraging historical context, enabling safer and more consistent navigation. It supports scalable deployment in real-time systems, improving trust and operational safety in autonomous driving.
Market Size (TAM)
$20–50B TAM for autonomous vehicle perception and planning software; $2–10B SAM from urban autonomous vehicle manufacturers and fleet operators. Driven by increasing urban deployment and safety regulations.
Potential Customers & Pain Points
- Autonomous vehicle manufacturers – Need robust perception under sensor degradation
- Fleet operators – Require safer navigation in dense urban traffic
- Automotive suppliers – Demand efficient real-time planning modules.
Business Model
Licensing the RCT-AD framework as a software module to autonomous vehicle OEMs and fleet operators, with options for customization and ongoing support.
Competitive Landscape
- Waymo
- Tesla Autopilot
- Aurora Innovation
- Mobileye
- Cruise
Implementation Challenges
- Integration complexity with existing autonomous driving stacks
- Real-world validation under diverse urban conditions
- Competition from established autonomous driving software providers
Validation Strategy
- Benchmark performance on public datasets like nuScenes
- Pilot integration with autonomous vehicle platforms in controlled urban environments
- Collect real-world operational data to refine reliability scoring and planning modules
Research Paper Overview
A Reliable Context-Aware and Temporal Planning Framework for Autonomous Driving
Summary
This paper presents RCT-AD, a framework that improves autonomous vehicle safety and planning stability by modeling feature quality and temporal consistency in perception. It selectively retains reliable sensor data to reconstruct degraded observations and uses a temporal planner for smoother, safer trajectories. Tested on nuScenes, it enhances detection, segmentation, and motion prediction while maintaining real-time efficiency.