SiMa.ai in Robotics

From Perception to Action, On-Device.

SiMa.ai enables a new class of intelligent robots — from autonomous mobile platforms to humanoid systems — with real-time on-device AI inference under 10W.

Get Started Today

Customer and Partner Stories

Why Customers Choose SiMa.ai for Robotics

Purpose-built silicon and software for intelligent robotic systems — from autonomous mobile platforms to humanoid robots.

Best-in-Class Performance per Watt

Industry-leading FPS with multimodal compute at the edge — enabling 8–10x improvement over incumbent GPU-based solutions.

Autonomous Mobile Robots (AMR)

Multi-modal compute for SLAM, object detection, and path planning — all on a single low-power chip enabling fanless, compact mobile platforms.

Humanoid and Collaborative Robots

Physical AI enables real-time perception and decision-making in humanoid robots and cobots, supporting safe human collaboration on the factory floor.

SLAM and 3D Navigation

16-channel LiDAR input at 10 FPS combined with 6-camera object detection at 30 FPS — simultaneous localization and mapping at the edge.

Conversational HMI

Run small language models like Gemma 3 locally on robot hardware — enabling natural language interfaces for operators without cloud connectivity.

Stereo Depth and Sensor Fusion

LightStereo depth estimation at 333 FPS fused with LiDAR, radar, and IMU data — accurate real-time 3D perception for safe navigation in dynamic environments.

Full-Stack Integration

How Modalix MLSoC Enables Robotics

One platform spanning every layer of the Robotics stack — from silicon to high-level applications.

Simulation and Debugging Tools

Simulate / Visualize / Debug

Data Generation

ISAAC SIM / Gazebo

Digital Twin

ISAAC SIM

Live Visualization

RViz

Analyze & Debug Live / Recorded Reality

Foxglove

Other Helper Tools

Analyze / Test / Manage

Performance Estimator

Fault Injection

Log Handling

Digital Twin / Fleet Mgmt

Customer and Partner Applications
Applications
HMI / UX
Micro services
SiMa.ai APIs
GStreamer
Python
C++
RoS2 Middleware
Libraries and Plugins
Depth Estimation
SLAM
Pose Estimation
Segmentation and Classification
VIO
ML/AI Accelerator
CNN to GenAI
Computer Vision
Multimedia
Video Enc/Dec
Sensors and Control
BSP: Kernel Drivers for Peripherals
OS (Yocto / Debian eLxr Pro®)
Hardware Product
Modalix™ MLSoC
Modalix™ MLSoC
System on Module (SoM)
System on Module (SoM)
PCIe Accelerator Card
PCIe Accelerator Card

Accelerated ROS Nodes

Frames / Second
333FPS

Stereo Depth Calculation

LightStereo

30FPS

Stereo Depth Calculation

SGM/SGBM

425FPS

Feature Point Detector

SuperPoint

126FPS

Feature Matcher

SuperGlue

429FPS

Loop Closure

NetVlad

770FPS

Detection / Segmentation / Pose

YoloV8

Applications

Autonomous Mobile Robots

Autonomous Mobile Robots

Autonomous Mobile Robots

We Enable

  • Real-time SLAM and obstacle avoidance on-device
  • Multi-sensor fusion: LiDAR, cameras, and IMU on one chip
  • Compact, fanless platforms for indoor and outdoor deployment

Why SiMa.ai

16-channel LiDAR at 10 FPS and 6-camera detection at 30 FPS — under 5W — enabling mobile robots that run all day on a single charge

Humanoid Robots

Humanoid Robots

Humanoid Robots

We Enable

  • Foundation model inference (GR00T) at the edge
  • Real-time multi-limb coordination and environment perception
  • Fast onboarding of new manipulation skills via Palette SDK

Why SiMa.ai

Edge AI compute powerful enough to run humanoid foundation models locally — no cloud latency, no connectivity dependency

Cobots and Manipulation

Cobots and Manipulation

Cobots and Manipulation

We Enable

  • Sub-10ms perception-to-action loop for precise grasping
  • Safe human-robot collaboration with real-time presence detection
  • Offline operation in air-gapped manufacturing environments

Why SiMa.ai

Vision AI running fully on-device — enabling cobots that adapt to dynamic environments without cloud round-trips

Outdoor and Field Robotics

Outdoor and Field Robotics

Outdoor and Field Robotics

We Enable

  • Wire-free navigation in unstructured outdoor environments
  • Robust vision models across variable lighting and weather
  • Cost-effective silicon for consumer and commercial platforms

Why SiMa.ai

Low-power edge AI that fits battery-constrained outdoor platforms — proven in production robotic lawn mowers and field inspection robots

Conversational HMI

Conversational HMI

Conversational HMI

We Enable

  • On-device LLM inference for operator guidance and Q&A
  • Voice and text interfaces without cloud connectivity
  • Hands-free control in noise-sensitive or safety-critical zones

Why SiMa.ai

Small language models like Gemma 3 running on SiMa.ai MLSoC — enabling natural language HMI entirely at the edge

Warehouse Automation

Warehouse Automation

Warehouse Automation

We Enable

  • Goods-to-person picking with vision-guided manipulation
  • Fleet coordination and dynamic path replanning
  • Inventory counting and anomaly detection without cloud

Why SiMa.ai

Scalable edge AI across heterogeneous robot fleets — single software stack, consistent performance per platform

Use Cases and Demos

Real-world robotics AI capabilities powered by on-device inference.

Autonomous Mobile Robots (AMR)

How It Works

SiMa.ai powers AMRs with multimodal compute for simultaneous localization and mapping (SLAM), object detection, and path planning — all on a low-power edge platform.

Key Benefits

  • 16-channel LiDAR input at 10 FPS for precise mapping
  • 6-camera object detection at 30 FPS for safe navigation
  • Low-power profile enabling fanless, compact mobile platforms
Brief