PITTSBURGH — Seegrid Corp. today announced that its vision-guided vehicles have traveled 2 million miles at customer sites without a single personnel safety incident. The company’s self-driving industrial vehicles for materials handling reached 1 million miles safely in April 2018.
In comparison, the U.S. Occupational Safety and Health Administration (OSHA) reported nearly 40,000 serious accidents and one fatality per week related to manually driven forklifts last year. Seegrid stated that its vision-guided vehicles (VGVs) are designed to “excel at routine and repetitive tasks, work seamlessly alongside human co-workers, and mitigate safety risks.”
Seegrid’s autonomous mobile robots (AMRs) use its proprietary Seegrid Vision navigation technology and redundant safety systems. Stereo cameras, algorithms, and machine learning enable the VGVs to collect and process more data about the environment than laser-based systems, the company said.
With major customers in manufacturing, distribution, and e-commerce order fulfillment such as Amazon, General Motors, Whirlpool, and Jaguar Land Rover, Seegrid claims to be a leader in autonomous materials handling.
The Robot Report spoke with Jeff Christensen, vice president of product at Seegrid, about the company’s milestone and its approach to autonomous materials handling.
Is the direct comparison between Seegrid’s robots and forklift usage and the accident rate reported by OSHA correct? Have there been near-misses?
Christensen: Manned industrial trucks move similar-sized loads, and “forklifts” is the colloquial term. They’re analogous.
Since our trucks can be driven manually, we see all sorts of shenanigans when people drive — they run into racking and overhead conveyors. When the vehicles are in autonomous mode, that doesn’t happen at all.
This is the message for those who don’t think automation is for them: Safety is the No. 1 reason for customers to change processes from manned to unmanned trucks.
Are automated systems safer in general, or is Seegrid a clear leader among them?
Christensen: Automated systems are safer in general than human drivers, assuming that a reputable vendor is satisfying all safety regulations. However, it’s one thing to say that you have a robot that’s safe, but it’s in a trade-show booth, not in production use, or you have five units in a pilot.
Lots of startups in this space may claim a perfect safety record, but we take great pride in the execution and demonstration of that safety through 2 million production miles.
Old-school AGVs [automated guided vehicles] have been around for a long time. In the fully autonomous space, we have a real lead in legitimate work at customer sites, where we’ve moved 8 billion to 10 billion pounds of materials. Our AMRs have driven themselves based on inference from the environment in real time, with zero safety incidents. That’s impressive in the real-world marketplace.
Can you give an example of a real-world environment where Seegrid’s products are being used?
Christensen: Every one of those 2 million miles has been in dynamic environments. In manufacturing, think automotive OEMs [original equipment manufacturers], Tier 1 suppliers, or tire manufacturers.
In an e-commerce or distributor facility, nothing is standing still. All of our experience has been in full-bore production plants or warehouses, where the robots are adjacent to manually driven vehicles, pedestrians, and moving products.
How have Seegrid’s vehicles improved with all the data you’ve been gathering?
Christensen: From an inward-facing perspective, we have data from a lot of different places that we put into a machine learning model. We’ve taken our navigational reliability, which was already the best, and made it better, based on the data we’ve collected from gritty, dirty environments that are hard at work.
On the customer-facing side, Seegrid has enabled them to do continuous improvement on their own. Our customers can apply this technology to an application or a job — such as taking parts and bringing them to the assembly line, or taking away finished goods — which we design out with them before they buy anything.
But once you start using something, you should see opportunities for improvement, following Toyota’s lean manufacturing model. A lot of customers see how they can apply our technology, making small tweaks to an autonomous route, which can always run the same way, unlike with a human driver.
They can retrain that small segment of a route on a truck themselves without disrupting production or calling us in. A shutdown and scheduling a vendor visit are points of friction that make incremental change seem too costly.
In addition, customers can replicate changes they like to the whole fleet. The route becomes a controllable variable in the operational materials flow, and the technology encourages them to create a virtuous cycle. Our biggest customers are proponents of doing exactly that.
Why do you think Seegrid’s vision-based approach is superior to alternatives?
Christensen: It’s all about sensor coverage — the more you can see, the more data you have, and the more you can do in software than be constrained by hardware. You have to be smart in terms of interpreting the data, but first you need that data.
You have to gather as much environmental data around the vehicle as possible in real time. Lidar is typically single plane — two to two-and-a-half dimensions. That’s the entirety of a mobile robot’s understanding, and then it must use that data and make decisions.
Conversely, we have hemisphere around and above the robot with pretty dense cloud of points, and a smaller portion of the data is moving versus static. It’s not really about lidar or cameras.
Cameras do give us a far wider, comprehensive view of the world all around, so you can make better decisions, especially in dynamic conditions. If you’re constrained by a sensor, everything else downstream is going to be constrained, and navigational reliability and safety will be poor.
What software or hardware refinements do you expect or are you working toward next?
Christensen: We’re certainly not resting on any laurels. We see a vision of interoperability for end-to-end materials flow, getting closer to full automation. That’s not lights-out operations, but automated handoffs or end-to-end flows without human touches.
There are lots of vendors tackling different parts of the market. There’s a wide variety of chassis types and automation.
We want to work with more truck models and keep our industry leadership in navigation. We want interoperability across all those types and then the software layer on top of that optimizing in real time.
Customers don’t think in terms of vehicles but in terms of what they have to move. That’s where it’s going in the long range — interoperability of load handling, manipulation, and sophisticated fleet management.
We didn’t actually start counting safe miles driven when Seegrid was founded back in 2003. We started in 2013, and it took five years to get to our first million miles. It took 500 days to get to our second million.
We’ve joked about having an office pool for the third million, and it wouldn’t surprise me if it’s less than a year. We have lots of customers who are now over the edge of realizing that the need for automation is real and that the technology is real.
The Robot Report has launched the Healthcare Robotics Engineering Forum, which will be on Dec. 9-10 in Santa Clara, Calif. The conference and expo focuses on improving the design, development and manufacture of next-generation healthcare robots. Learn more about the Healthcare Robotics Engineering Forum.