We make robots see with radars
on-road, off-road, and beyond.
Perception, no matter what.
LiDAR-like performance. Fraction of the cost.
Any radar. Any circumstances.
Introducing world’s first
dedicated AI for radar perception.
Most automated vehicles, robots and systems today are using LiDARs, even though they are expensive, fragile, and struggle in harsh conditions such as rain, night, fog, dust, smoke, and so on.
Perciv AI solves all these problems by replacing LiDARs using AI-driven radar and radar-camera fusion software. Combined with affordable, off-the-shelf radars, our technology enable scalable and robust autonomy.
Compatible with systems on-road or off-road, indoor or in-air
Perciv AI’s products has been deployed on cars, trucks, forklifts, off-roaders, drones, and beyond.
We help any robot to understand their environment regardless the conditions.
On-road applications
ADAS and Smart City use-cases
Indoor applications
Logistics, warehouse,
and Smart home use-cases
Off-road applications
Agriculture and Defense use-cases
In-air applications
Collision Avoidance, take-off and landing, and C-UAS/drone detection use-cases
Perciv’s product line contains solutions for the four main
challenges every robot faces:
Localization: Placing the robot in the 3D world -> Perciv EgoTrack
Detection: Sensing and detecting any objects / obstacles around the robot -> Perciv FreeRoad
Classification: Classifying and tracking objects around the robot -> Perciv Objects
Fusion: Increasing accuracy and robustness by sensor fusion ->Perciv Fusion
Learn about all four products below!
Perciv EgoTrack:
Radar-only ego-motion estimation
Perciv EgoTrack provides reliable ego-motion and odometry information purely based on one or multiple radars.
Highly useful for GPS-denied areas, such as urban canyons or parking garages, and provides reliable odometry even off-road.
Radar-only ego-motion estimation
No GPS / IMU, or other sensor is needed
All-weather performance
Translational drift as low as 2% (1% with IMU)
Perciv FreeRoad:
Radar-only free-space Estimation
Perciv FreeRoad provides reliable occupancy information about the environment purely based on radar sensors, detecting both dynamic and static objects regardless of their classes. Very useful feature for e.g. lost cargo and collision avoidance challenges.
Reliable, radar-only occupancy
All-weather, day and night
Detects all static or dynamic objects
Obstacle detection regardless of class
Up to 150 m, up to 99.99% Recall for VRU-s
Perciv Objects:
Radar-only object detection and classification
Perciv Objects is a state-of-the-art radar-only object detection, classification and tracking module, providing an object list with quality never seen before for radars. It detects cars, pedestrians, and two-wheelers out of the box - additional classes are available on demand.
Object list with 3D pose, velocity, class, and track
All-weather, day and night
Lidar-Level classification performance
Existing or on-demand object classes available
#1 performance on the View-of-Delft and ManScenes benchmark datasets
4x times better classification than leading tier-1’s solution
Perciv Fusion:
The strength of radars and cameras
Every product of Perciv AI is available in radar-only and in radar + RGB/Thermal camera fusion versions.
Fusing light sensors with radars increase classification accuracy and redundancy at a system level, truly challenging the performance of LiDAR sensors at a fraction of the cost.
Early- and late-fusion both offered, along with multi-level fusion solutions, depending on the exact needs.
Flexibility from the core
With Perciv, the perception pipeline adapts to you, not the other way around.
Sensor-agnostic: From low-end to high-end radars, Perciv AI works with any kinds of radar hardware.
Compute-agnostic: by using extremely light-weight, < 2 TOPs networks, Perciv is compatible with a wide range of compute solutions.
Modular and Customizable: Unlike Tier-1s, Perciv offers its features in a modular way - each module configurable and fine-tunable to your use-case right out of our SDK.
AI-Driven: Our foundational AI models are trained on large amount of heterogeneous data, and are fine-tuned to your exact application, mounting, and use-case.
State-of-the-Art Radar AI Research
The Team of Perciv AI has been pushing the state-of-the-art radar perception for years.
With over 600+ citations, we established our place in the cutting edge or radar AI research.
You can read a couple of our latest papers on the right.
-
LeAP: Consistent multi-domain 3D labeling using Foundation Models.
Read it here -
Multi-class Road User Detection with 3+1D Radar in the View-of-Delft Dataset
-
3+1D Radar and Camera based Point Cloud Estimation for 3D Road User Detection
-
CNN Based Road User Detection Using the 3D Radar Cube
-
Occlusion-aware sensor fusion for early crossing pedestrian detection.
-
DPFT: Dual Perspective Fusion Transformer for Camera-Radar-based Object Detection.
Read it here -
A Deep Automotive Radar Detector using the RaDelft Dataset
Partners
News
