Acceleration Robotics builds Robotic Processing Units and FPGA/GPU IP cores that offload ROS and ROS 2 workloads to heterogeneous hardware, delivering sub‑microsecond latency, deterministic timing, and lower power consumption for safety‑critical robots. Their ROS‑compatible accelerator library and custom design services let OEMs and integrators speed up perception, transform, and communication pipelines without extensive software changes.
Funding
Funding not disclosed
Founders
Product
Problem
Robotic applications that rely on ROS or ROS 2 often run on general‑purpose CPUs, leading to high latency, limited determinism, and excessive power consumption. This hampers real‑time performance for autonomous mobile robots, industrial manipulators, and other safety‑critical systems.
Solution
Acceleration Robotics provides custom hardware acceleration platforms—Robotic Processing Units (RPUs) and FPGA/GPU‑based IP cores—designed to map ROS computational graphs efficiently onto heterogeneous compute resources. By integrating multiple CPUs, a GPU, and programmable FPGA fabric, the solutions deliver faster execution, deterministic timing, and lower energy use. The company also offers a suite of ROS‑compatible IP blocks that accelerate networking, perception, and coordinate‑transformation pipelines, enabling robots to meet real‑time constraints without extensive software rewrites. Consulting and IP‑design services help customers tailor these accelerators to specific robot workloads, shortening development cycles and improving return on investment.
Target Audience
Primary customers are robotics OEMs and system integrators developing autonomous mobile robots, collaborative robot arms, industrial manipulators, and healthcare robots that rely on ROS or ROS 2 for control and perception.
Features
- Multi‑core architecture combining up to 16 CPUs, a GPU, and an FPGA on a single board for heterogeneous acceleration.
- ROS and ROS 2 API‑compatible IP cores that offload perception, transform (tf2), and communication stacks (UDP/IP, DDS‑RTPS) to hardware.
- Ultra‑low‑latency networking: ROS 2 messages processed in <2.5 µs and round‑trip latency under 11 µs, achieving >60× speedup over CPU‑only implementations.
- Deterministic execution through FPGA‑based eager evaluation, providing consistent timing for safety‑critical control loops.
- Power‑efficient design that reduces energy per operation compared to pure CPU solutions, extending robot runtime.
- Modular IP library allowing customers to integrate specific accelerators (e.g., collaborative collision avoidance, perception pipelines) into existing ROS workspaces.
- Consulting and custom IP design services to adapt acceleration blocks to diverse robot platforms and application domains.