Planning
Stretch 5
Lessons, the landscape, technology trends, the AI roadmap and architecture for the next generation.
Stretch's competitive moat is its AI ecosystem, not hardware. With 216+ GitHub stars on stretch_ai, the largest mobile manipulation research community, and integration into Open X-Embodiment, Hello Robot is the default platform for home robot AI research. Stretch 5 should double down by upgrading compute (Jetson Orin) and making the AI developer experience frictionless: the best-integrated AI platform for home manipulation research.
Strategic priorities
- Jetson Orin compute upgradeEnables VLA models, future-proofs AI stack
- SAMD51 + dedicated I2C busesEliminates critical firmware reliability risk (I2C reentrancy)
- ESP32-S3 with defined UART protocolClean WiFi/BLE/OTA architecture
- eFuse power protectionSafety + FCC compliance
- FCC Class B design-inRequired for consumer market
- Enhanced Dex Teleop + LeRobot pipelineData flywheel is the long-term moat
Competitive landscape
| Platform | Company | Price | Status | Threat | Differentiator |
|---|---|---|---|---|---|
| Stretch 3 | Hello Robot | $24,950 | Shipping | low | Lightest, most affordable research platform |
| TIAGo | PAL Robotics | ~$80K+ | Shipping | low | Full humanoid torso, industrial-grade, ROS |
| Mobile ALOHA | Stanford (open-source) | ~$30K BOM | Research | high | Bimanual, learning from demonstration focus |
| Unitree G1/H1 | Unitree | $16K–$90K | Shipping | high | Humanoid, legged locomotion, Chinese supply chain |
| Figure 02 | Figure AI | N/A | Pilot programs | medium | General-purpose humanoid, massive VC backing |
| 1X NEO | 1X Technologies | N/A | Pre-production | medium | Home humanoid, backed by OpenAI |
| LeRobot | HuggingFace (open-source) | Open | Active | high | Standardized LfD framework, could commoditize Stretch's role |
Technology trends
Single neural networks that take images + language instructions and output robot actions directly. π₀ (Physical Intelligence) and RT-2/RT-X (Google DeepMind) represent the leading edge.
π₀ (pi-zero)RT-2 / RT-XDiffusion Policy
For Stretch 5: Stretch 5 compute must support 3B+ parameter VLA models at >5 Hz inference. Requires dedicated GPU (Jetson Orin).
AI/ML integration roadmap
Near-Term (6–12 months)
- Object GraspingNow: OWL-v2 detection → heuristic graspWith foundation models: VLA model (π₀-style) generates grasp trajectories end-to-end
- Task PlanningNow: LLM prompt → fixed operation sequenceWith foundation models: LLM with affordance grounding — plans only feasible actions
- NavigationNow: A* / RRT on voxel mapWith foundation models: Learned navigation policies for dynamic obstacles + social norms
- Voice InteractionNow: Whisper STT → GPT-4o → Piper TTSWith foundation models: On-device multimodal model (Gemma/Qwen) for low-latency, private interaction
Medium-Term (12–24 months)
- Dexterous Manipulation via LfDNow: Manual programmingWith foundation models: Dex Teleop + Diffusion Policy — learn to fold laundry, load dishwashers, sort items
- Autonomous ExplorationNow: Pre-mapped navigationWith foundation models: DynaMem + LLM-guided exploration ('go find all dirty dishes')
- PersonalizationNow: Static configurationWith foundation models: Fine-tune to individual homes — learn object locations, user preferences, layouts
- Multi-Robot CoordinationNow: Single robot onlyWith foundation models: Two Stretch robots with bimanual Dex Teleop for complex tasks
Long-Term (24+ months)
- World ModelsNow: Reactive control onlyWith foundation models: Predictive models of home physics — enables planning without trial-and-error
- Embodied QA & MonitoringNow: No passive monitoringWith foundation models: 'Is the stove off?' 'Did grandma take her medicine?' — always-on home awareness via VLMs
Compute requirements for AI stack
| Model class | Example | Parameters | Min VRAM | Speed |
|---|---|---|---|---|
| Vision encoder | SIGLip-so400m | 400M | 2 GB | 30+ FPS |
| Object detector | OWL-v2 large | 300M | 3 GB | 10+ FPS |
| Segmentation | SAM2-base | 90M | 2 GB | 15+ FPS |
| LLM (local) | Qwen2.5-7B | 7B | 6 GB | 20 tok/s |
| VLA model | π₀-small | 3B | 8 GB | 5+ Hz |
| Full stack | Perception + LLM + VLA | — | 12–16 GB | Pipelined |
Recommendation: Jetson Orin NX 16 GB as baseline, Orin AGX 64 GB for flagship. This future-proofs Stretch 5 for the VLA model wave.
- V1 PIMU: Added INA228 power monitor, pre-charge circuit, safe PC shutdown, system shutdown mode (103µA vs 12–30mA), brake button float mode, IMU+Mag, better USB hub ICs
- Charging improved: 10A (2.5hrs vs 4.5hrs)Requires 36V 8A adapter specifically
- BMS comms limited to 9600bpsBlocking I2C at 1kHz was problematic
- EMC: USB hub renumeration from ESD fixed with new ICsGrounding still critical
- Stretch 3 passed Class A EMC in both operational and charging modes
Development roadmap
- 1Architecture & Schematic
- Define power architecture (Efuse, INA228 alerts)
- Consolidate 3V3 rails
- Master/slave UART protocol spec
- Select Class B EMC-friendly USB hub ICs
- Select Jetson Orin NX vs AGX for compute
- Design SAMD51 with dedicated I2C buses (IMU/Mag/INA228 separated)
- 2Prototype & Bring-up
- In-house pre-compliance EMC scans
- Actuator protection validation
- Non-blocking I2C implementation + DMA transfers
- Reverse current path testing
- Jetson Orin integration + AI stack validation
- ESP32-S3 UART protocol bring-up
- 3Validation & Compliance
- FCC Class B formal testing
- Automated RDK test suite complete
- SOC accuracy validation (INA228 vs BMS)
- System grounding audit
- VLA model inference benchmarks on Orin
- Dex Teleop + LeRobot pipeline validation
- 4Production Readiness
- Final BOM review & cost optimization
- Manufacturing test fixtures
- Firmware OTA update pipeline (via ESP32-S3)
- Documentation & handoff
- 'Contribute data' toggle for fleet learning