While deep learning methods have proven useful for autonomous navigation, the approach requires vast amounts of data and computing power, and produces solutions that are inefficient, fragile and unaffordable for many classes of applications. Opteran’s “Natural Intelligence” approach, based on the function of insect brains, provides an alternative for generating complex, autonomous behaviors for machines.
Editor’s Note: Robotics Business Review‘s coverage emphasizes innovation, including start-up companies (or ‘young’ companies). RBR “Start-Up Profiles” highlight individual start-up companies using a consistent, templated format that makes for quick, yet informed reading, that also simplifies comparative analysis.
If you are interested in having your start-up profiled in Robotics Business Review (RBR), please contact Dan Kara (dkara[AT]wtwhmedia.com. There is no charge for submitting and posting Start-Up Profiles on RBR.
Opteran – Company Profile
In 2022, Opteran will seek Series A funding to continue investment in research, product development and commercialization.
By mimicking how insects use optic flow, Opteran can complete tasks such as seeing, sensing objects, obstacle avoidance, navigation and decision making in a much more transparent and compliant way compared to deep learning methods.
Opteran – Offerings Profile
Primary Target Markets – 1) Consumer, 2) Logistics / Supply Chain, and 3) Transportation
Technology / Product / Service(s) – Opteran is developing a small, lightweight, and ultra-low power silicon-based approach to autonomy called Natural Intelligence. Based on eight years of insect brain research by James Marshall and Alex Cope, Opteran is reverse-engineering insect brains onto silicon chips to produce high performance algorithms requiring no data center, no neural-net, or extensive pre-training.
By mimicking how insects use optic flow, Opteran’s modular autonomy platform can complete tasks such as seeing, sensing objects, obstacle avoidance, navigation and decision making in a much more transparent and compliant way compared to deep learning methods. The board weighs approximately 30g, draws less than three watts of power, and operates at the network edge only, without training, and is regulatory compliant – opening the device up a wide range of use cases.
The Opteran autonomy platform consists of:
- Opteran See – 360 degrees of electronically-stabilized vision using low cost 2D CMOS cameras.
- Opteran Sense – Bio-inspired optic flow estimation, state estimation, collision detection and avoidance.
- Opteran Direct – Simultaneous Localization and Mapping (SLAM) that works like brains do, robust to the problems faced by classical feature extraction approaches whilst also being much more computationally efficient.
- Opteran Decide – Action selection for autonomy, that functions as brains do without the need for rule-based decision making or deep reinforcement learning.
According to Opteran, this is only the start on the path to enable full natural autonomy in machines by re-engineering how insect brains have evolved in the natural world to both understand and interact with their environments.
Value Proposition – By re-engineering 600 million years of evolution in insect brains onto a silicon chip, Opteran is developing an alternative to Deep Learning neural networks. This approach enables machines to see, sense, navigate and make autonomous decisions with a smaller, lighter and ultra-low powered solution that operates at the edge of the network without extensive pre-training.
Demand Drivers – Organizations wanting to automate, streamline and increase the intelligence of their robots, or autonomous vehicles and manufacturing processes have long been interested in artificial intelligence as a means to deliver dramatic performance increases. However, existing deep learning models are often not effective in the real world and require too many resources (compute power, training data, servers, data scientists) to deploy on the edge, limiting the widespread adoption of autonomous machines.
Opteran’s technology transforms the use case for commercial, industrial and consumer robotics, as it will enable real-time, edge only autonomy on a wide variety of autonomous vehicles including tracked, wheeled, legged and aerial. For example, in an early trial the team was able to control a sub-250g drone, with complete onboard autonomy, using fewer than 10,000 pixels from a single low-resolution panoramic camera.
Opteran licenses its Intellectual Property directly to OEM robotic or vehicle manufactures for integration into systems.
Opteran – Business Model(s) and Competition
Business Model – Evaluation of the technology is via the Opteran Development Kit (ODK) which facilitates integration and testing of the technology. Opteran licenses its Intellectual Property directly to OEM robotic or vehicle manufactures for integration into systems. Alternatively, Opteran can work with its partners to provide pre-integrated silicon solutions, as required.
Partners – THis information is confidential at this time.
Customers – The company cannot reveal its customers at this time. They are working with a range of companies in the drone (rotor and co-axial), wheeled, tracked and legged autonomous mobile robots (AMRS) including healthcare, warehousing, construction, mining and other. The company is also working with companies producing consumer robotics such as self-balancing autonomous bicycles.