Netradyne Providing Deep Learning Solutions For Effective Fleet Management

Interaction Netradyne Deep Learning Effective Fleet Management
Netradyne Providing Deep Learning Solutions For Effective Fleet Management

Shared mobility is increasingly gaining prominence world over. Such a scenario necessitates an efficient management of various vehicle fleets for providing a systematic solution for end-customers. Among the various fleet management solutions offered, an important focus area is leveraging technologies like Artificial Intelligence (AI) and Deep Learning that can pave the way for an efficient deployment of vehicles.

An efficient fleet management can bring in higher profitability for companies that own large vehicle fleets. Netradyne offers vision-based Deep Learning solutions for vehicle fleets in monitoring or alerting drivers and fleet managers to facilitate best and safe utilisation of vehicles and drivers. Auto Tech Review caught up with Teja Gudena, Vice President, Engineering – Devices, Netradyne, to know more about its flagship product platform – Driveri, back-end technologies and future mobility solutions that can be deployed in fleets.

BACKGROUND

Netradyne was established in 2015 by two former Qualcomm employees, with an intent to build AI and Deep Learning-based products for smart surveillance or automotive applications. Market research revealed that the auto industry has a big business use case for the development of vision-based products. Netradyne set up technology innovation centres in California and Bengaluru, with teams working towards creating an AI platform to disrupt commercial vehicles and driver analytics.

Gudena said Netradyne typically has three types of customers -transportation fleets, ride share fleets and OEMs. Transportation fleets are further divided into commercial and corporate fleets, wherein the latter refers to fleets owned by companies for the transportation of their own inventories. The company is working with commercial transportation fleets in the US and is conducting advanced trials with all types of Indian customers, he added.

DRIVERI

Driveri is built on an NVIDIA Tegra X1 platform. It features four cameras with one each for inward and outward recording and two for side-angle view recording. It is a fully-connected device over LTE, and provides fleet managers with information that helps them analyse and anticipate different scenarios drivers might face. This also helps fleet managers not just in rewarding drivers but also providing proper guidance and coaching to them.

Netradyne brings in a strong differentiation via Driveri in the sense that videos captured by the front and inward-facing cameras are processed on the device itself in real-time, said Gudena. This is the first device that can process videos from two cameras (inward and outward-facing) at the same time on the device itself, he noted. Cameras usually record and store videos in their memory, or consequently upload these videos on cloud storage. In both cases, there are limitations that the footage can be viewed or used for analysis only after an accident or incident has occurred.

In terms of real-time processing, the device has all vision-based Deep Learning capabilities to process the video and infer any specific driver behaviour use cases, he explained. For example, the inward-facing camera can detect the movements and gestures of a driver that make the system conclude that the driver is sleepy. This information is instantly uploaded to the cloud and the fleet manager is alerted about the same so that corrective measures can be promptly initiated.

Gudena said that the complete Deep Learning IP for the Driveri platform is developed in-house. The platform comprises three components – device & software; deep learning algorithms and cloud back-end. These components are completely developed by Netradyne, save for the device manufacturing that is carried out with its manufacturing partners in Gurgaon and China.

Driveri is currently installed in the US as a passive feature, wherein no information is relayed to the driver but transferred directly to the fleet manager. This is because drivers are more mature and sensitive to being monitored, explained Gudena. However, the company is conducting trials with Indian customers, who have put in request for alerts in the form of audio feedback to the driver in case of situations like drowsiness or violations. The platform enables fleet managers to evaluate the driving behaviour and assign scores to each driver, called Green Zone Score. In India, Netradyne has received more demands for inward-facing camera for fleets with real-time alert to the driver, he added.

The Driveri device also has the capability to connect to the CAN BUS, which enables the reading of various details of engine data and process such data for further analysis. This would provide information on the optimum method of driving to ensure better performance and efficiency as well as help in predictive maintenance that leads to an efficient utilisation of fleet vehicles.

In terms of ADAS, while there are a number of use cases, a few are more relevant to India, as in the case of collision warning, Gudena said. There is the capability of sending a possible upcoming collision alert to the driver. He added that the next level of ADAS requirement from Netradyne’s OEM engagements is in sending out a collision warning on to the vehicle itself, for it to take corrective action. These active systems would reduce accidents and ensure safer forms of mobility.

CONCLUSION

Large Tier 1 firms are investing in safety and autonomous driving technologies, providing Netradyne a favourable proposition. There is a good possibility to collaborate with such companies, where Netradyne complements its existing set of core deliverables, added Gudena.

TEXT: Naveen Arul