Autonomy · Phase 1 in execution

Self-driving, built for three wheels.

EMC Squared is building a bespoke edge-AI driving foundation tuned to the exact kinematics of a Delta three-wheeler — so the trike that delivers your order today can drive itself tomorrow, with no new hardware.

Book a demo See the roadmap

A mature stack, aimed at one hard problem.

Rather than rebuilding autonomy from scratch, EMC Squared stands on a production-grade software base and spends its engineering on what's genuinely new: control and perception for a three-wheel platform.

275 TOPS

NVIDIA Jetson Orin AGX

Edge compute with the headroom to run six HD camera streams at once — and scale into LiDAR fusion later with no hardware swap.

6 cameras

Spatial perception array

A surround camera stack delivering deterministic, end-to-end perception for low-speed urban environments.

<100 ms

Perception latency

Target end-to-end latency across the full camera array, validated on edge hardware before anything touches a public road.

Production-grade architecture

The stack runs on the Autoware.Universe framework, natively optimized through NVIDIA TensorRT. That frees the team to focus on proprietary control algorithms and tricycle kinematics instead of foundational middleware.

Decoupled, capital-efficient testing

By separating multi-tensor perception validation from high-speed actuation, EMC Squared iterates software fast, keeps the physical platform out of harm's way, and preserves capital at every step.

Phase 1: the Momentum Gate.

A strict four-stage validation pipeline that de-risks perception and control before public-road deployment.

STAGE 01

Static tensor validation

Locking in deterministic perception across the six-camera array and tracking the distance of static obstacles inside a 10-meter radius.

Active

STAGE 02

Dynamic shadow mode

Tracking moving obstacles at 5–10 mph with zero tracking drops, proving the system handles live urban motion on edge hardware.

Next up

STAGE 03

Closed-loop actuation

Wiring validated perception into the drive-by-wire system to trigger autonomous, low-speed automatic emergency braking.

Queued

STAGE 04

Geofenced autonomy alpha

Full autonomous path-following and cornering inside a bounded 25×25 m zone — proving control under three-wheel lateral dynamics.

Queued

The path to driverless delivery.

Five phases, each one generating real data and de-risking hardware on the way to a scalable commercial fleet.

PHASE 1In execution

Closed-loop ADAS validation

Validate the camera perception stack and low-speed drive-by-wire actuation.

PHASE 2Next

Sensor fusion & autopilot

Integrate LiDAR for HD-map localization and comprehensive Level 4 capabilities.

PHASE 3Planned

Fleet beta & shadow telemetry

Deploy a limited test fleet to harvest human-driven field data and train edge-case models.

PHASE 4Planned

Regulatory validation

Complete ISO-compliant safety cases and secure local operating permits for driverless deployment.

PHASE 5Goal

General availability

Launch commercial, fully unmanned three-wheelers in USA markets.

Want to follow the autonomy program up close?

Book a session to see the platform, the perception stack, and where the roadmap goes next.

Learn more

EMC Squared's autonomy platform is under active development. Stage, phase, and performance figures describe an in-progress engineering roadmap and development targets, not the current capabilities of a publicly available driverless vehicle.