Telecom Techniques for Autonomous Sensing

A full-stack autonomous sensing platform for ten percent the cost of conventional methods is what Perceptive promises. Their path to this dramatic cost reduction is to apply both principles and components of modern telecommunications. Similar to a modern wireless network, fiber optic cables carry laser and radar signals between small, inexpensive antennas and a central processing unit. Cameras help guide their sensors to focus on what is important.

Besides cost, Alberto Stochino, Ph.D., Perceptive Founder &CEO, outlines many other advantages to this technique compared to a traditional integrated antenna/processing approach, including:

• Because of the availability of raw data and multiple sensor inputs at the processing unit, they can effectively synthesize sensors via software. As such, their central unit can make decisions that would not be possible when each sensor is independent.
• Improved reliability through a reduction in components and the placement of those components in a more secure, protected location within the vehicle, as compared to the exterior of the vehicle where sensors are located today.
• Reduced power consumption because of the reduced component count.
• It scales as they are using off-the-shelf components from telecommunications and data centers.
• Hardware upgrades would presumably be easier as only the central unit should need replacing as sensors/processing improves.

Stochino sums up their design philosophy, “You don’t want to fight the laws of physics. So, you need to think about something that is not a brute force approach…..the good thing is you don’t have to.”

Interview Highlights #

Click on the timecode to open a new window at that point in the video.

• 00:42 – Alberto explains that they felt a new approach would be required to meet cost, reliability, and mass deployment objectives for autonomous vehicles.
• 02:38 – A high-level block diagram of their system is shown. Stochino emphasizes that their focus is driverless (SAE Level 4/5) vehicles.
• 06:30 – Stochino explains that bringing multiple sensor inputs to one location allows a holistic approach so that the sensors work together to make smarter decisions. They combine laser, radar, and cameras. He says “When you intersect them (the sensors), you can make much stronger conclusions.”
• 08:05 – Centralizing improves reliability. He indicates that one of the advantages is that they share components through multiplexing. Stochinco indicates that they have remapped the telecommunications approach to sensing. It is analogous to wireless architectures, like 5G.
• 11:17 – Stochino suggests that it will be an order of magnitude cheaper than traditional approaches. He expects pricing to drop further, as the price/performance of components, such as silicon photonics, continue to improve.
• 12:42 – Even though they were only founded in 2018, they already have a vehicle outfitted with their technology.
• 13:45 – Perceptive provides a sensing platform that gives the developers access to the raw data. The data may be used for training or inference and Perceptive lets the user decide where to process the data. Perceptive also provides consulting services to help the developer optimize the sensing platform for a given application.
• 16:09 – Stochino expects pilot programs, measured in tens of vehicles, towards the end of 2020. He envisions providing a hardware platform, although the level of integration will vary depending upon the customer.
• 18:26 – Stochino discusses his background, which includes stints at Apple, NASA, and LIGO.
• 21:05 – He believes their approach applies to autonomous aircraft, as well as other sectors, such as agriculture, shipping, and trucking
• 23:01 – Focusing on what is important – the signal, that is – reduces the processing power requirements. Perceptive’s website indicates that “Cameras extract scene semantics and direct lidar and radar attention to specific objects/regions of interests.”
• 24:37 – Stochino comments on SpaceX’s Starlink system and the use of lasers as part of a mesh satellite-to-satellite outer space backbone.
• 26:45 – Will Lidar be eye-safe? Stochino indicates that one of the advantages of their approach is that they can run the lasers at lower power than conventional Lidar, providing additional safety margin compared to traditional approaches.
• 28:26 – Alberto doesn’t provide details, but indicates they have a unique way of scanning the horizon.