Industrial robotics is often associated with repeatable operations in controlled work cells. But many tasks in factories and other real-world environments involve irregular objects, changing surroundings or close interaction with people. Research into robotic manipulation, mobility and control is addressing these challenges, with the aim of helping robots operate more reliably beyond tightly constrained setups.
Several complementary areas of work illustrate what this progress requires: algorithms that interpret and act on the physical world, robot designs that can adapt to terrain, and control methods that support safe movement and collaboration. These are research directions rather than ready-made solutions for every production line, but they offer a useful framework for evaluating where automation may be extended.
Industrial robotics depends on more capable manipulation
Handling a rigid part with a known shape is different from manipulating a cable, garment or other deformable object. Flexible materials change shape as they are touched or moved, making their position and response harder for a robot to predict. This can complicate tasks such as grasping, routing or transferring an object from one location to another.
Research at the University of Michigan’s Autonomous Robotic Manipulation Lab focuses on algorithms that let robots interact with objects and collaborate with people. The work spans motion planning and manipulation, from developing efficient methods and examining their theoretical properties to testing them on real robots. It also brings together sensing and higher-level reasoning, so that a robot’s actions can account for information about its surroundings.
The research includes manipulation of deformable objects as well as grasping in cluttered environments. These topics are relevant to industrial settings where materials may not maintain a fixed shape or where objects are arranged unpredictably. The practical challenge is not just to produce a movement, but to give a robot methods for responding when the object or environment does not match a simple, preprogrammed scenario.
Robot mobility must account for real-world terrain
Automation systems are frequently designed around a defined workspace, but mobile robots may need to travel across surfaces that are uneven or obstructed. A robot that can maintain its movement over grass, slopes or other irregular ground has a different range of potential uses than one limited to flat, predictable floors.
One example in the research is BigANT, a modular robot developed in Shai Revzen’s lab. A redesign enabled it to walk over grass, climb hills and cross uneven terrain, including the undulating Wave Field on the University of Michigan campus. The work demonstrates the value of combining a robot’s physical design with its movement capabilities: navigating an obstacle is not solely a software problem if the machine itself cannot handle the terrain.
Modular construction also offers a way to fabricate different robot configurations using modules and stock materials. In research and prototyping, that approach can support the creation of robots for varied tasks. It does not eliminate the need to validate a robot for its intended environment, but it highlights how adaptable hardware can complement adaptable planning and control.
Control algorithms link movement to stability
Legged robots provide another example of the close relationship between hardware and control. Cassie Blue, a bipedal robot developed in Jessy Grizzle’s group, was designed to walk without a safety gantry and to withstand trips and falls. It also included a camera and additional processing capability. The robot began walking with the group’s control laws within six weeks of its arrival, according to the report.
For automation engineers, the broader point is that mobility requires more than actuators and a programmed path. Control algorithms must coordinate movement and balance, while sensing and computation help the system respond to what it encounters. The right combination depends on the application: a fixed industrial arm and a mobile legged robot face different operating conditions, but both rely on coordinated hardware, software and control.
Human-robot collaboration requires reliable interaction
Robots intended to work alongside people need to function in environments where human activity is part of the task. Research on manipulation and motion planning therefore includes human-robot collaboration, with attention to interacting safely and reliably in human environments. This is especially important when a robot must handle objects while people share the workspace or when tasks cannot be fully isolated from human judgment.
Work on walking exoskeletons provides a related example of control methods being applied to a device that moves with a person. Grizzle’s team worked with the company Wandercraft on algorithms for walking motion and a gait suited to the exoskeleton’s design. The project involved fitting the software into the device’s control package and testing the algorithms on the physical system. The collaboration illustrates the importance of translating models and control methods into real hardware, particularly when people are directly involved.
Another research direction addresses how robots and people exchange information. A method inspired by the game of 20 Questions was designed to present a set of questions in advance, while accounting for the fact that some answers may be more certain than others. The researchers identified possible applications in information gathering and communication between robots and other computers. For industrial systems, the general lesson is that effective interaction depends not only on physical movement but also on how information is requested, represented and used.
What these research directions mean for automation planning
For plant teams and system integrators considering broader robot use, these examples point to several practical evaluation questions:
- How variable is the task? Flexible objects, clutter and changing surroundings may require more capable sensing and manipulation than a fixed, repeatable pick-and-place operation.
- What must the robot move through? Terrain and obstacles influence both the mechanical design and the motion-planning approach.
- Where will people be involved? Shared tasks put additional emphasis on predictable movement, reliable control and effective information exchange.
- How will performance be validated? Algorithms need to be integrated with sensing and tested on real hardware and representative tasks, not assessed in isolation.
These questions do not imply that every production application needs a mobile or highly adaptive robot. They help identify where conventional automation may be sufficient and where variability, mobility or human interaction makes additional capabilities relevant.
Conclusion: match robot capability to task complexity
Progress in industrial robotics depends on coordinating algorithms, sensing, mechanical design and control. Research into deformable-object manipulation, uneven-terrain mobility and human-robot interaction shows how robotics can address tasks that are more complex than simple, repetitive motion. For technical decision-makers, the practical implication is to define the real operating conditions first, then assess whether a robot’s hardware and software can handle them reliably. Careful task analysis and real-world validation remain central to choosing an appropriate automation approach.