Autonomous Driving HD Map Versioning for Over-the-Air Updates

Overview of Technical Issues:

During over-the-air map updates, the version control module insufficiently manages the transition between old and new map data, causing the positioning module to retrieve inconsistent or partially updated map information, which leads to localization errors and compromised autonomous driving safety during the update process; the goal is to enable seamless map version updates without interrupting vehicle operation or degrading positioning accuracy.

Solution directions generated for this problem

Problem Direction 1 :

ImproveMap version synchronization speed
VS
ConstraintSystem computational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #10 Preliminary action
Cross-domain applicability Assess applicability
Fact checking method and system utilizing format
Innovative Solution Refine solution

Idle-period map pre-staging with atomic pointer switching for zero-latency version transition

Pre-stage new map during vehicle idle time with atomic pointer switch
How to solve :
  • Detect vehicle idle states (parking, charging, speed <5 km/h) via CAN bus
  • trigger background map download and validation using spare CPU cycles (≤10% load threshold)
  • store new map in pre-allocated shadow memory buffer (reserved at boot, 512MB fixed allocation)
  • perform CRC32 integrity check and tile-level validation during idle periods (acceptance: 100% tiles pass checksum, positioning test error <0.15m)
  • when update completes, execute atomic pointer swap in <100ms by flipping memory base address register — positioning module instantly accesses new map without data copying or real-time processing
Expected Effect : Sync time reduced from 5-15s to <100ms; CPU load during switch <5%; memory overhead fixed at 512MB
Risk Control :
  • idle period detection accuracy insufficient
  • shadow buffer corruption during pre-staging
  • pointer swap atomicity failure under multi-thread access

Problem Direction 2 :

ImproveData state detection precision
VS
ConstraintVersion control module complexity

Inspiration 1 : Cross-domain reference

Application Principle: #26 Copying
Cross-domain applicability Assess applicability
Method and apparatus for generating virtual software platform based on component model and validating software platform architecture using the platform
Innovative Solution Refine solution

Lightweight metadata tagging for tile-level version validation

Embed metadata tags in map tiles for validation
How to solve :
  • Attach a 16-byte version tag (8-byte timestamp + 8-byte CRC64 checksum) to each map tile header during OTA packaging
  • positioning module reads tag via direct memory access to determine tile validity without traversing complex state structures
  • Implement single-pass validation logic: compare tile timestamp against update epoch threshold and verify checksum match — binary valid/invalid decision in <5 CPU cycles per tile, eliminating multi-layer indexing and state machines
  • Deploy ring buffer tile cache (512 tiles, ~50MB) storing only tags of active-zone tiles within 500m radius
  • update tags atomically via compare-and-swap instruction during version switch, achieving <100ms transition with zero architectural overhead
Expected Effect : Detection precision: tile-level; module complexity: +8% code; CPU overhead: <2%; memory: +50MB; transition time: <100ms
Risk Control :
  • timestamp synchronization drift across distributed tiles
  • checksum collision probability in high-frequency updates
  • cache invalidation race condition during atomic swap

Problem Direction 3 :

ImproveSystem operational reliability
VS
ConstraintSystem computational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #11 Beforehand cushioning
Cross-domain applicability Assess applicability
Variable-geometry timepiece display mechanism with resilient hand
Innovative Solution Refine solution

Pre-validated shadow map buffer with atomic pointer switching for zero-downtime updates

Pre-validate map in shadow buffer during idle
How to solve :
  • Allocate a shadow map buffer (300MB) at system boot
  • download and validate new map tiles in background during vehicle idle periods (parking, charging) using CRC32 checksum verification
  • Perform integrity pre-verification in shadow buffer: tile completeness check, coordinate continuity validation, and cross-tile boundary consistency test — all completed before vehicle operation resumes
  • Execute atomic pointer switch (<10ms) when positioning module queries map: flip memory pointer from active buffer to pre-validated shadow buffer via single compare-and-swap instruction, eliminating real-time verification overhead
Expected Effect : Positioning failure rate 0%, switch time <10ms vs 5-15s baseline, CPU spike eliminated
Risk Control :
  • shadow buffer validation incomplete during unexpected departure
  • memory pointer race condition during switch
  • tile corruption undetected in background validation

Problem Direction 4 :

ImproveVersion control module complexity
VS
ConstraintSystem computational resource consumption

Inspiration 1 : Cross-domain reference

Application Principle: #2 Taking out
Cross-domain applicability Assess applicability
Exploiting input data sparsity in neural network compute units
Innovative Solution Refine solution

Sparse version index with non-zero delta extraction for map updates

Extract only changed map tiles for version tracking
How to solve :
  • Implement sparse indexing structure that stores only modified tile coordinates and checksums (typically 5-15% of full map), eliminating full-map version tables — reduces memory overhead from 200-300MB to 30-50MB
  • Deploy delta extraction engine during background download: compute tile-level hash comparison, generate compact change list with tile_ID + version_tag (16 bytes per tile), store in lightweight lookup table with O(1) access via hardware hash acceleration
  • Position module queries only active tiles within 500m radius against sparse index — unmodified tiles bypass version check entirely, cutting CPU validation cycles by 85-92% while maintaining tile-level granularity
Expected Effect : Memory overhead -83%, CPU load -87%, <100ms switch time maintained
Risk Control :
  • hash collision in sparse index
  • delta computation accuracy during poor network
  • tile boundary synchronization errors
Patsnap Eureka Solution