HomeResourcesBlue River Technology Developers Iterate 2.5x Faster to Empower Farmers

Blue River Technology Developers Iterate 2.5x Faster to Empower Farmers

TL;DR

Blue River Technology is on a mission to solve monumental agricultural challenges by empowering farmers with intelligent machinery so they can operate more productively and sustainably. One of the systems that Blue River designed powers a line of autonomous See & Spray™ John Deere tractors. The ability to process AI computer vision workloads is a core capability of the tractors, which are used for tillage and weed control.

By adopting the Anyscale Platform to run Ray on AWS, Blue River cut regression testing time for AI workloads in product development by more than half. With this improved productivity, Blue River now provides faster insights from computer vision images and rolls out new product updates sooner.

Overview

Computer vision drives Blue River Technology’s smart tractors for more efficient and targeted crop management.

Blue River Technology is at the forefront of designing smart machinery that helps farmers automate time-intensive tasks, capture better data about crop yields and efficiencies, and optimize the amount of inputs they need to purchase.  

The company's automated tractors — sold as autonomous See & Spray™ John Deere tractors — can operate with minimal supervision, for more hours, and in varying weather conditions. AI computer vision is crucial to the tractors' autonomous systems. While working, a tractor uses six pairs of stereo cameras for a 360-degree view that detects all objects in the field. After the system captures the images, it sends them through a deep neural network that classifies each image within about 100 milliseconds and determines if the tractor needs to stop, depending on the objects it encounters and the tractor’s location. The tractor also continuously checks its position for accuracy based on a pre-defined geofence. These capabilities mean the autonomous tractors are not only highly controlled, but they also can differentiate between crops and weeds for targeted and cost-effective herbicide application.

The Challenge

Time-consuming regression testing holds back productivity.

Blue River's fast and nimble culture encourages engineers to talk directly with farmers about their challenges and then develop and iterate on solutions rapidly. But Blue River's development team faced a bottleneck in its regression testing suite. It took an entire day to process the image and video data collected by the autonomous tractors — around 132 database files. The delay constrained the team's productivity by limiting the number of experiments it could conduct.

Among the most time-consuming steps was the conversion of data from Blue River's proprietary database format to the ROS bag format, which lasted seven to nine hours. Another concern was how long it took to generate a report after the conversion and playback. This step would process sequentially, taking three to four hours.

The Solution

Anyscale Ray scales parallel processing to remove AI workload bottlenecks.

Blue River needed to remove inefficiencies related to processing large amounts of data in real time and move to a system that could scale as needed. The solution was to migrate the AI workload processing to Ray managed by Anyscale to tap the power of parallel processing at scale. The new architecture consisted of an Anyscale Ray cluster running on AWS to manage the Ray platform in Python, allowing Blue River to parallelize the conversion of its file formats to ROS bag format as well as its report processing. 

The Impact

After adopting Ray and Anyscale, Blue River Technology experienced dramatic productivity improvements to its product development processes and can now pass those productivity gains on to its customers.

  • Higher productivity with the same team: Regression testing time and report processing sped up by more than 2x, making the engineering team 2.5x more productive.

  • Resource efficiency: The team can now use shared hardware to move faster and more efficiently. 

  • Reliability: The Anyscale Platform automatically retries jobs and provides notifications.

  • Fail-safe operations: The AWS cluster removed many single points of failure, making the testing process more robust and predictable.

By mitigating the testing bottleneck, Blue River Technology not only accelerated its development cycle, but also improved the reliability and efficiency of its autonomous tractor technology. These improvements are putting modern productivity and sustainability tools in the hands of farmers faster — all driven by the AI that powers Blue River's pioneering intelligent machinery.

Ready to try Anyscale?

Access Anyscale today to see how companies using Anyscale and Ray benefit from rapid time-to-market and faster iterations across the entire AI lifecycle.