LiDAR Point Cloud Analysis for ADAS Perception Systems
Analyze LiDAR point clouds to debug ADAS perception pipeline failures
What You Can Do
You can rapidly isolate whether perception failures originate in hardware, preprocessing, detection models, or sensor fusion logic by analyzing 3D spatial data and coordinate system transformations. This skill helps you debug object detection edge cases, validate multi-sensor alignment, review confidence scores, and diagnose environmental-specific performance drops that impact autonomous vehicle safety decisions.
Features
identify sensor noise, dead zones, reflectivity issues, and hardware failures from raw data
analyze confidence scores, false positives/negatives, and edge case failures (occlusion, weather, low reflectivity)
troubleshoot multi-LiDAR fusion and coordinate system misalignment between LiDAR, camera, and vehicle frames
validate downsampling, filtering, outlier removal, voxelization, and grid representation conversions
identify performance degradation patterns tied to specific weather conditions or scenarios
compare detection outputs against sensor data to verify labeling accuracy
trace failure points across hardware, preprocessing, detection models, and fusion stages
Example Output
Example 1: Occlusion Detection Failure Analysis
Input: Point cloud data from highway scenario where vehicle was not detected behind truck
Output:
- ✓ Identified 40% point density reduction in occluded zone
- ✓ Confirmed detection model confidence dropped below 0.3 threshold
- ✓ Validated ground truth label matches sensor capability limits
- ✓ Recommended increasing confidence threshold tolerance for post-processing
Example 2: Sensor Alignment Issue
Input: Multi-LiDAR array data with inconsistent detections
Output:
- ✓ Detected 2.3° yaw misalignment in rear unit
- ✓ Showed 0.15m translation offset in coordinate frame
- ✓ Generated calibration adjustment matrix
- ✓ Provided fusion weighting recommendation for misaligned unit
What's Included
- SKILL.md instruction file with systematic analysis methodology:
- Point cloud anomaly assessment checklist (sensor failures, preprocessing artifacts, confidence issues):
- Coordinate system transformation reference guide for LiDAR-camera-vehicle frame conversions:
- Multi-sensor fusion validation framework (alignment, timing, confidence weighting):
- Edge case analysis template (occlusion patterns, weather impact, reflectivity effects):
Who It's For
- ADAS/Autonomous vehicle perception engineers debugging detection failures
- Sensor fusion specialists troubleshooting multi-LiDAR integration issues
- Data validation engineers validating ground truth labeling and detection confidence
- Perception systems engineers investigating environmental-specific performance drops
- Hardware integration engineers diagnosing sensor alignment and calibration problems
Best For
- Debugging failed object detections in edge cases and environmental conditions
- Validating point cloud preprocessing pipelines (filtering, downsampling, voxelization)
- Analyzing multi-sensor alignment and coordinate system transformations
- Reviewing detection confidence scores and false positive/negative patterns
- Investigating perception pipeline failures across hardware, preprocessing, and model stages
You might also like
You can systematically analyze suspension geometry, load paths, and component interactions to validate design decisions and identify optimization opportunities. Claude acts as your collaborative design partner to map force distributions, explore how geometry changes affect performance characteristics, compare suspension architectures, and reconcile conflicting engineering constraints in independent double-wishbone, multi-link, and other suspension systems.
You can transform vague customer NVH complaints into actionable diagnostic protocols by systematically mapping vibration sources, transfer paths, and receiver responses. The framework guides you through symptom normalization, operational condition analysis, and isolation strategy development—reducing diagnostic iteration cycles and accelerating root cause identification across powertrain, chassis, interior, and aerodynamic domains.
You can pinpoint the exact source of sensor fusion failures—whether they stem from misaligned camera/LiDAR/radar extrinsic calibration, timestamp desynchronization, data validation gaps, or algorithmic issues. This skill decomposes multi-modal perception errors into sensor-specific contributions and generates quantifiable diagnostic reports that directly guide calibration and testing teams toward fixes rather than blind troubleshooting.
You can efficiently diagnose complex NVH concerns by working through a structured framework that correlates customer symptoms with potential powertrain failures. This skill helps you narrow diagnostic possibilities, recommend targeted testing sequences, distinguish between component failures and tuning issues, and substantiate warranty claims with technical evidence.
You can model transient and steady-state thermal behavior in EV power electronics, calculate junction temperature rise during driving cycles, and design cooling strategies that meet automotive thermal specifications. This skill helps you size heat sinks, evaluate thermal interface materials, compare active versus passive cooling approaches, and validate designs against AEC-Q200 and ISO 16750-3 standards.
You can systematically analyze thermal performance of battery pack designs using heat generation physics, thermal resistance calculations, and boundary condition modeling. The skill helps you identify cooling system inadequacies, assess failure modes like temperature runaway and hot-spot formation, and generate optimized design recommendations across passive air, active liquid, hybrid, and advanced cooling strategies for cylindrical, pouch, and prismatic cell formats.
You can decode SAE J2012 standard CAN diagnostic trouble codes (P, B, C, and U codes) with manufacturer-specific extensions, analyze multi-frame fault sequences and temporal patterns across multiple modules, and generate structured diagnostic hypotheses that differentiate between electrical, communication, and algorithmic failures. This skill transforms raw DTC data into actionable investigation vectors, reducing diagnostic analysis cycles from hours to minutes while improving root cause accuracy.
You can transform raw accelerometer data, sound pressure measurements, and frequency spectra into structured diagnostic insights. Claude helps you organize test data, interpret Campbell diagrams and waterfall plots, quantify source contributions through transfer path analysis, and correlate physical measurements with customer complaints. This systematic approach replaces intuitive troubleshooting with hypothesis-driven test plans and targeted solutions.