List view
Hybrid cloud-edge connectivity and Azure Arc integration — closing out the roadmap with edge-local autonomy, MQTT-based communication, and Arc-managed lifecycle. Key deliverables: • Arc automation — Arc-connected cluster lifecycle management (#281), Arc GitOps configuration for edge fleet (#308) • Azure IoT Operations — AIO integration for device management and telemetry (#284) • Edge-local capabilities — edge-local inference without cloud connectivity (#294), local model cache with version management (#297), offline-first operation mode (#298), edge-initiated training with federated aggregation (#300) • MQTT communication — MQTT broker deployment for robot-to-cloud messaging (#301), MQTT topic hierarchy and schema design (#306) • Observability — edge telemetry pipeline with local buffering (#307), distributed tracing across edge and cloud (#309) • Documentation — hybrid architecture decision record (#290), hybrid deployment guide (#304), edge-cloud connectivity test suite (#303) This lighter milestone closes the roadmap with buffer for spillover from earlier milestones. It establishes the hybrid operating model where robots function autonomously at the edge with intermittent cloud connectivity, managed centrally through Azure Arc.
Due by April 25, 2026•4/15 issues closedSecurity hardening, human-machine interface (HMI), and deployment automation — making the system production-ready and operator-accessible. Key deliverables: • Security hardening — secrets management with Azure Key Vault integration (#192), TLS certificate automation for service endpoints (#196), network policy enforcement across namespaces (#197), security audit automation and compliance reporting (#198) • HMI operator interface — operator dashboard for monitoring and control (#193), emergency stop integration and safety interlock UI (#194), multi-robot fleet status view (#195) • Deployment automation — Terraform module for multi-region AKS (#268), automated cluster scaling policies (#269), blue-green deployment pipeline (#270), canary release infrastructure (#271), GitOps with Flux CD (#272), environment promotion workflow (#273), infrastructure drift detection (#274), automated rollback on health check failure (#275), deployment telemetry dashboard (#276), resource quota management (#277), cost allocation tagging (#278), multi-tenant namespace isolation (#279), deployment runbook automation (#280), disaster recovery failover (#283), SLA monitoring and alerting (#288), incident response automation (#292), post-incident review tooling (#293) Security is prerequisite for production deployment — secrets, certificates, and network policies must be in place before exposing services. The HMI layer transforms the platform from a developer tool into an operator-ready system. Deployment automation ensures repeatable, safe rollouts across environments.
Due by April 18, 2026•3/17 issues closedEdge-side data recording and upload pipeline — capturing high-fidelity robotics demonstration data on physical hardware and reliably transferring it to the cloud for training. Key deliverables: • Recording pipeline — demonstration recording with metadata (#200), realtime bag recording unit tests (#203), chunked bag recording for large ROS2 datasets (#204), ROS2 bag compression with zstandard (#205) • Recording triggers — GPIO-driven recording trigger with demonstration mode (#201), position-based trigger for automated demonstrations (#202) • Upload infrastructure — SQLite upload queue with priority scheduling (#206), resumable upload to Azure Blob Storage (#218), bandwidth throttling configuration (#209), upload telemetry and status reporting (#214) • Testing and benchmarks — recording pipeline integration tests (#208), upload queue integration tests (#210), chunking and compression benchmarks on Jetson Orin (#215) • Documentation — edge data collection and recording architecture (#199), chunking and compression configuration guide (#207) This milestone completes the edge-to-cloud data flow: demonstrations recorded on physical robots flow through the recording pipeline, get chunked and compressed for efficient transfer, and upload reliably to Azure Blob Storage where the cloud pipeline (v0.8.0) ingests them for training (v0.9.0).
Due by April 11, 2026•0/14 issues closedUnified training and evaluation pipeline — from job submission through model assessment. Training and evaluation are tightly coupled: every training run produces checkpoints that feed directly into evaluation. Key deliverables: • Training SDK — Python SDK for job submission (#243), CLI wrapper (#244), YAML schema for job configuration (#240), metrics logging callback (#241), Azure ML compute and environment setup template (#245), SDK unit and integration tests (#246) • Checkpoint management — automated checkpoint saving to blob storage (#248, #251), training job auto-retry on failure (#249) • Training documentation — monitoring and troubleshooting guide (#247) • Evaluation framework — policy rollout engine for offline evaluation (#260), policy inference loop in Isaac Sim (#256), checkpoint selection and A/B comparison (#252), evaluation report export (#250) • Evaluation infrastructure — Isaac Sim Azure VM deployment template (#259, #262), robot URDF loading and environment configuration (#261), unified evaluation metrics aggregation service (#265) • Evaluation UI — evaluation dashboard web UI (#266), CSV export for dashboard data (#263), dashboard integration tests (#267) • Domain evaluation — ROVE task-level domain evaluation module (#328) Delivers a complete train-evaluate-compare loop that enables data-driven model selection before deployment.
Due by April 4, 2026•0/22 issues closedCloud-side data ingestion, conversion, and validation pipeline — the infrastructure that transforms raw robotics data into training-ready datasets in Azure. Key deliverables: • Ingestion infrastructure — Event Grid trigger for blob upload detection (#219), storage account folder structure and lifecycle policies (#238), OneLake shortcut configuration (#239) • Data conversion — ROS bag to LeRobot parquet conversion module (#230), conversion pipeline Bicep/Terraform template (#231), frame validation event-driven post-processing (#232), integrate validation into conversion pipeline (#234) • Quality validation — frequency analysis for frame gap detection (#220), validation report schema and generator (#226), validation test suite with synthetic data (#227), vision-based frame quality validation (#237) • Pipeline orchestration — Azure Fabric pipeline orchestration (#221), conversion pipeline architecture documentation (#229) • Security and isolation — BYOK encryption configuration (#233), RBAC configuration for multi-tenant isolation (#236) This milestone establishes the cloud data backbone that training (v0.9.0) and edge capture (v0.10.0) both depend on for data flow.
Due by March 28, 2026•1/15 issues closedComplete dataviewer application — a real-time visualization and diagnostic tool for robotics data streams. Delivers the full stack from backend services through frontend UI to containerized deployment. Key deliverables: • Build and scaffolding — build and lint remediation (#381), scaffolding and dev experience (WS-12 #393), containerization (WS-17 #398), logging migration (WS-19 #400) • Backend architecture — service architecture (WS-13 #394), middleware stack (WS-05 #392), input validation (WS-06 #391), detection service hardening (WS-07 #390), storage adapter improvements (WS-08 #389), backend test quality (WS-16 #397), exception handling (WS-18 #399) • Frontend — app shell and build (WS-15 #396), component quality (WS-14 #395), state management (WS-10 #387), hooks quality (WS-11 #386), API client consolidation (WS-09 #388) • Platform features — platform requirements and resource templates (WS-001 #414), LLM backend orchestration (WS-002 #413), CUDA/driver compatibility layer (WS-003 #412), agent response validation (WS-004 #411) • Auth Phase 2 (WS-20 #401) Each work stream is a tracking issue that decomposes into sub-tasks, enabling parallel development across the full application stack.
Overdue by 5 day(s)•Due by March 21, 2026•18/40 issues closedGovernance formalization and documentation restructuring to achieve OpenSSF Best Practices Passing badge and establish a comprehensive, navigable documentation structure. Key deliverables: • OpenSSF Best Practices — Passing badge achievement (#96), README badge addition (#93) • Contribution governance — DCO/CLA adoption (#97), vulnerability reporter credit policy (#103), deprecated interface policy (#105), reused component update process (#104) • Project continuity — access continuity plan (#100), bus factor mitigation (#101) • Code quality policies — enforce strict warnings (#99, #106) • Documentation restructuring (FR-series) — YAML frontmatter for 9 files (#347), glossary (#350), architecture overview (#351), simplified root README (#352), slim CONTRIBUTING.md to pointer (#357), operations hub with troubleshooting (#358), monitoring and cost management guides (#359), reference hub with script docs (#360), workflow reference docs (#361), terraform-variables.md (#364), environment-variables.md (#365), architecture diagrams (#366) This milestone transforms the project from a working codebase into a well-governed, well-documented open source project that meets industry standards for security and maintainability.
Overdue by 12 day(s)•Due by March 14, 2026•3/14 issues closedFoundational release establishing the CI/CD pipeline, Python tooling, and code quality infrastructure that all subsequent milestones depend on. Key deliverables: • Release automation — fix release-please tag creation guard (#174), remove harden-runner from workflows (#137) • Python dependency rationalization — eliminate sys.path hacks via editable installs (#191), rationalize legacy requirements.txt files (#189), add per-workflow Dependabot entries (#190) • Unified code coverage — Codecov integration (#422), Python CI upload (#423), codecov.yml config (#424), pyproject.toml coverage settings (#425), Pester threshold alignment (#426), orchestrator workflow updates (#427) • Security scanning — Bandit static analysis workflow (#88), dependency-review GHSA allowlist audit (#310) • Documentation tooling — automated ms.date freshness checking (#164) • Platform fixes — devcontainer missing dependency (#179), NAT Gateway availability zone support (#182) • Isaac Sim 5.x upgrade to fix env.close() shutdown hang (#374) • PowerShell CI helpers — Test-PowerShellVersion and null-guards (#337) • HVE-Core workflow migration tracking (#52)
Overdue by 19 day(s)•Due by March 7, 2026•6/8 issues closed