As the autonomous machine industry is gaining momentum and self-driving car companies like Alphabet Inc.’s Waymo and Uber are competing for the spotlight in showrooms, the sensors in a car that scan its surroundings are becoming increasingly intelligent.
Today, one software developer says its product can understand the human mind.
Perceptive Automata, run by a team of experts in machine learning, vision science, human cognition, and artificial neural networks, among other things, seeks to help driverless cars and robotic systems in predicting the behavior of pedestrians and other moving objects.
Perceptive Automata is building a “human intuition machine,” the startup’s co-founder and chief executive officer, Sid Misra, said in an interview with VatorNews.
Autonomous cars today have a pretty good sense of object detection and recognition, whether it’s a person, a car, or an inanimate object. They also know the edges and markings on the road, and the locations of objects – how far away, in which direction, and how fast they are moving. However, when it comes to anticipating a pedestrian’s, a cyclist’s, or a human-driven vehicle’s next step, technology has fallen behind, Misra said.
Supplied with a system from Perceptive Automata, a machine would make judgments about a pedestrian’s intention and predict whether a person on the side of a road plans to cross, or is waiting for a bus, Misra explained. Such a system would be helpful in autonomous driving, driver assistance for human-driven cars, and other robotic applications.
“The artificial intelligence space has seen a lot of improvement in the last five years because deep learning has become much more powerful and effective,” Misra said.
“Deep learning, a type of machine learning, initially had success due to the rapid increase in its ability to apply computational resources to large datasets. However, at Perceptive Automata, we use a different approach to deep learning that lets us remain very computationally light while capturing a uniquely rich characterization of the human response. We’ve been doing research on our approach for a very long time, we caught the lead, so we are continuing to build on that.”
Last week, the four-year-old Boston company announced a partnership with Renovo’s AWare system, a mobility software technology. In its statement, Renovo said that with Perceptive Automata it has gained access to “critically important algorithms to enable the safer deployment of automated vehicles and a smoother ride experience for passengers of automated mobility services.”
Misra said about the partnership, “We’re very excited about that. This partnership exemplifies how we think the ecosystem of the autonomous vehicle space would evolve.”
A complex problem such as the creation of an intuitive machine for the safety of driverless cars cannot be solved without help from the best experts in the autonomous industry, he said. Together with its partners, Perceptive Automata plans to begin testing its system’s abilities under various circumstances.
In 2017, the startup scored a funding of $4 million from First Round Capital and Slow Ventures, as well as other institutional and Angel investors from deep machine learning and automotive backgrounds.
The funding, which Misra said would last the company for another four to six quarters, would be used to demonstrate the performance of Perceptive Automata’s system running real-time in a car and reliably predicting the behavior of cyclists and pedestrians.
Then, upon the completion of the initial testing period, Misra said, “We can more publicly show how our software enables [machines] to be safer, more comfortable, and predictive for people inside and outside the car.”
Image: Perceptive Automata
