Scaling mobile robots for success


AMR deployments have been gaining ground to address labor availability issues. Fleets often start small in pilots and add robots over time, but how do you get to a better system? Success factors include proper solution scoping, RaaS options, integration platforms, and may involve software to orchestrate multiple AMR fleets.

By Roberto Michel · November 10, 2022

More automation doesn’t necessarily mean better automation. That’s why an autonomous mobile robot (AMR) deployment with an impressive number of bots in the fleet isn’t necessarily more effective than one with just a handful.

While it’s common for AMR deployments to start small and get bigger, success with AMRs takes more than growing the fleet. You need to consider how quickly and easily you can add more robots and people into a robotic-enabled workflow to meet spikes in demand, software capabilities, and the ability to integrate with other systems as well as human labor.

AMRs come in various form factors and payload capacities. Some larger AMRs are designed to transport heavier loads, assistive-pick AMRs work in concert with human pickers, and goods-to-person (GTP) solutions bring goods to picker workstations. Use of AMRs aren’t mainstream yet, but with warehouse operators struggling to find enough hourly workers, AMRs are catching on because many operations simply can’t count on securing enough human labor.

“The challenge is that there aren’t enough people to fill those roles any more, or they don’t want them, all while the cost of labor is increasing,” says John Santagate, vice president of robotics and voice for Körber Supply Chain, a provider of supply chain execution software. “As a result, organizations are looking at how to reduce the increasing costs, and also the risk associated with not being able to have enough labor, so they are looking at mobile robots.”

AMR solutions are often designed to work alongside people, reducing the non-value-added walking or “travel” involved under traditional methods, especially when software wasn’t used to optimize pick paths.

Using pilots or multiple implementations of AMRs, small tweaks to the system can gain further efficiencies.

For example, with assistive-pick AMRs, the human picker stays in a compact zone, and a succession of bots swing by to guide the worker through picks with minimal travel. That brings ergonomic and associate satisfaction factors into the mix when assessing what constitutes success.

“We’ve seen incredible results from our customers in terms of productivity and throughput, but also with some of the softer benefits, like improved employee happiness and retention,” says Kaitlin Peterson, senior director of product marketing for Locus Robotics, an AMR provider. “They are walking less, by more than half, which reduces fatigue, and pushing heavy picking carts, which carries the risk of strains or injury, is eliminated with our solution.”

Scope then scale

There is no ideal size for an AMR fleet. For instance, many Locus customers have around 50 bots in fleet, but some work fine with as few as a dozen. On the other hand, one customer runs 500 robots at a site that has picking on a four-level mezzanine, notes Peterson.

Figuring out how many AMRs are needed for an operation comes down to proper assessment of an operation’s order mix, its existing processes and layout, service levels and other characteristics, says Santagate.

“We look at their demand volume, their historical demand, we look at SKUs and dimensionality, and layout and configuration of their warehouse, like where would the induct and drop off points to a system be,” says Santagate. “All of those nuances impact the level of robotic capacity that’s going to be required and which approaches will support peak volumes. You should also consider how to possibly reshape your processes to make better use of a robot-enabled picking environment.”

The throughput improvements with AMRs can be impressive. Integrated Supply Network, an automotive tool and equipment distributor that uses a Körber warehouse management system (WMS) integrated to Locus Robotics’ solution, gained a 266% improvement to picking productivity with the AMRs versus its previous method.

Most AMR implementations are measured in weeks, not months, says Santagate, while robotics-as-a-service (RaaS) arrangements from AMR vendors offer a way to scale up the hardware needed to handle peaks without upfront capital expense. Another benefit to AMRs, Santagate says, is that user interface features on the bots like touchscreen tablets speed up onboarding. “You don’t need a week anymore to train someone to do picking—you can train someone in about an hour, which is a critical element in the benefits of robotic-enabled picking,” he says.

Kristi Montgomery, vice president of innovation with Kenco Group, a third-party logistics provider (3PL), agrees that proper solution scoping is a key first step to any automation project, AMRs included. At some of its DCs, Kenco uses Locus Robotics’ AMRs to accelerate each picking, which Montgomery says has increased the lines per hour picked in those DCs from roughly 30 to 40 lines per hour to 120 to 150 lines per hour. “We were very deliberate about establishing the success metrics prior to the implementation,” says Montgomery.

The human element to system success

Workstation ergonomics is a critical consideration in the success of goods-to-person systems.

Getting advanced warehouse automation and robotics configured and deployed effectively spans multiple considerations, starting with proper solution scoping against the order mix and factors like item dimensions, integration to warehouse management system (WMS) software, and strategies for adjusting the system to peak demand.

With all these factors to consider, it can be easy to overlook how human workers will interact with and support systems, points out Colin Thompson, vice president of operations for Vanderlande USA, a global provider of warehouse automation solutions.

“It’s easy to forget the human side of the success of these systems, but it’s people who are configuring the software, operating the goods-to-person (GTP) workstations, maintaining the solution, so without intelligent human input every step of the way, the automation doesn’t really do anything on its own,” Thompson says.

Operations and maintenance managers, for example, need to reach consensus on the metrics that will govern a system’s priorities, and help set up the metrics to watch, says Thompson. People in maintenance need to learn how to run simple troubleshooting procedures. GTP solutions such as shuttles that bring goods to workers, call for close attention to worker ergonomics, he adds.

“One of the first things to consider when getting [GTP] workstations is to focus on ergonomics,” he says. “As a vendor, we spend time analyzing movements, putting motion sensors on people to collect data on where the strain points are, and that research goes back in the design of the workstations. Even small things, like anti-fatigue mats, can make a difference, and making height adjustments easy is important.”

Another key human factor with GTP solutions is to devise incentives that encourage positive competition or “gamification,” rather than punishment, since GTP automation generates detailed data on picking performance such as lines-per-hour picked, notes Thompson. “Metrics need to be used in the right way,” he says. “Your high performers want to be measured. They want to see how they’re doing against other people.”

A phased approach to AMRs can be a good idea, adds Montgomery, because it’s a learning experience. “Even in a pilot implementation, you typically aren’t going to learn everything that could possibly happen on the production floor, so it can be good to go about it in a phased approach—to bring in the robots, see how they function, see how people react to them, see how well they integrate to WMS, and see how process might change. Because inevitably, when you add automation, whether its mobile robotics or large automation projects, your processes might need to change a bit. Then once you have those processes well defined and worked in, then maybe it’s time to bring in other solutions or systems you might want to integrate [AMRs] with,” he says.

Top Notch Distributors, a wholesale distributor of architectural door hardware, has rolled out an assistive pick AMR solution from 6 River Systems at multiple sites, learning from each deployment. The mobile robots, which 6 River nicknames “Chucks,” were first deployed in Top Notch’s Carson City, Nev., facility, which allowed Top Notch to validate the technology in a lower-risk environment, says Patrick Houlihan, director of operations for Top Notch.

Mobile robotics solutions typically offer dashboards that allow managers to see fleet use patterns.

“Next, we implemented Chucks into our larger St. Charles, Mo., facility, where we saw improvements in the health and safety of our employees, lowered operating costs per order and standardized work processes across our entire network,” says Houlihan. “With improved efficiency and productivity in Carson City and St. Charles, we implemented Chucks in our Mansfield, Mass., facility. In comparison to previous methods, we’ve reduced associate training time by 75%, seen a 50% reduction in steps per day, and more than doubled productivity, leading to higher employee satisfaction.”

Working jointly, 6 River and Top Notch determined they could increase throughput by spacing out the order batch waves. At times, Top Notch started to delay inducting orders onto the AMRs by completing inventory replenishment tasks first thing in the morning, which led to greater productivity later in the shift, says Matt Fitzgerald, director of product management at 6 River Systems. “The time spent in replenishment led to efficiencies in the order picking process, allowing teams to complete picking work two hours earlier, despite the later start,” he says.


Many end user organizations integrate AMRs to additional systems to gain efficiencies. Most assistive-pick AMR solutions need to integrate to a WMS to obtain order data, while some implementations sites integrate AMRs to systems like automated conveyors or pack-out lines, or voice systems, especially for picking cases to larger format AMRs.

According to Fitzgerald, 6 River’s partner program supports integration to systems such as automatic box builders, autobaggers, and vertical lift modules.

In some cases, AMR solutions might integrate with a manufacturing execution system (MES) or an enterprise resource planning (ERP) solution. The need to simplify integration of multiple robotic solutions with WMS has given rise to new vendors, including SVT Robotics, whose software Kenco uses to streamline integration of robotics to its WMS.

Integration of Fetch AMRs from Zebra Technologies with fixed position scanning, as well as software integration to a MES-type system, has allowed Bespoke Manufacturing Co. (BMC), a U.S-based apparel manufacturer, to automatically capture detailed data on the movement of work-in-process (WIP) materials while automating materials moves with AMRs, according to J. Kirby Best, president and CEO of BMC.

Phoenix-based BMC worked with Zebra partner S&H Systems to devise the AMR solution, which went live in summer 2022. The solution was aimed at enabling WIP visibility to make better manufacturing decisions. According to Best, 378 scan points are currently in the process with a few handheld scan devices, but about 95% of the scan points feature fixed-position equipment, also from Zebra.

According to Best, a previous system used conveyor, but the AMRs were an attractive alternative for the new Phoenix site because they could save floor space. The fixed-position scanning, Best adds, is highly accurate compared to handhelds and also saves production staff time by automating nearly all data collection as WIP materials (garment pieces like sleeves) move through the process on one of 15 AMRs.

“We were either going to go with a conveyor belt system, or we were going to go with AMRs [for automated transport],” says Best. “We went with a conveyor system last time, and when we learned we could free up floor space with AMRs we went with the mobile robots.”

For Best, the solution makes use of the flexibility of AMRs, but overall, the intent was to create a manufacturing visibility solution that captures granular, real-time data about where work is to optimally route materials and, in some cases, reassign production workers to different stations using the production management software the AMR solution integrates with from a company called iCreateone, which also developed the front-end commerce platform and design collaboration software BMC uses.

At a strategic level, Best adds, the overall solution allows BMC’s Phoenix operation to be cost and time competitive with overseas apparel manufacturers by supporting fast, customized apparel manufacturing, tapping into the strong base of skilled sewers in the Phoenix area.

“In the long run, what the overall system is really doing is allowing us to challenge offshore manufacturing with what we think is a better product delivered faster,” says Best. “The idea of labor saving didn’t really enter my mind, though there are some efficiencies to the data capture. Primarily, we were interested in what we could do with the information being gathered, to load balance our operation.”

Multi-agent orchestration

It’s still relatively early days for mobile robots, but as more companies begin leveraging AMRs, it expected that more companies will have fleets from multiple vendors—perhaps some large format AMRs to move pallets or for case pick, assistive-pick AMRs for fulfilling another part of the order mix, or perhaps a GTP-focused fleet for another part of the order mix.

This heterogenous fleet trend is giving rise to software that can manage fleets from multiple vendors. “At the core of our software is the ability to work with different resources or ‘agents’ that are available, and that includes people,” says Samay Kohli, CEO at GreyOrange, an AMR solution provider that sees its “GreyMatter” software as tackling this need, which it refers to as multi-agent orchestration.

“We support multi-agent orchestration with what we call drivers,” Kohli explains. “We have different drivers for different classes of agents, including robot systems with different payloads, for example, or traditional automation hardware, or humans with different job roles or titles. You need software that can orchestrate this whole plethora of agents, and which is flexible and easy to use for a client or integrator.”

In addition to drivers, says Kohli, the software applies artificial intelligence (AI) and machine learning as a data science foundation, and delays making agent assignments until the last possible minute to consider resource availability and optimize assignments based on current conditions.

BMC, an apparel manufacturer, uses Fetch AMRs from Zebra Technologies to automate tote transport between sewing stations and other work-in-process steps.

“Fundamentally, you can’t have a rigid system, because robots may be charging or otherwise not available, sometimes there is heavier traffic in an area, the SKU mix changes, and of course the order pool changes,” Kohli says. “The system will determine what’s the best combination of work and inventory that can be allocated to resources, knowing the current state each resource is in.”

Others also recognize the need to orchestrate different systems. Santagate says Körber is developing a solution called its Unified Control System, which will integrate and control different working models—fixed automation, voice systems and AMR systems. The trick, he agrees, is to have a standard platform that is easy to work with.

“We get there by enhancements to the software architecture that can manage a heterogeneous or a multi-bot environment and still keep it on a standardized platform,” he says.

Read more from Modern Materials Handling 

Previous ArticleGreyOrange CEO Explains How and Where Mobile Robots Can Add Value  Next ArticleAMR multi-fleet orchestration software market growing
icon-angle icon-bars icon-times