Virtual Dynamic Simulations
Simulations that incorporate accurate physics and realistic interaction between robots and the environment is used for testing and validation of the research conducted in the lab. We are constantly working to improving and expanding the capabilities of these simulation software. High realism and accuracy provided by these simulations reduce the time it takes to implement an idea on physical hardware and improve research efficiency.
Robot Dynamics and Control
One of the main research areas of the BAU Robotics Lab is advanced control methods based on force/torque control that enable safe interaction of robots with humans and the environment. The complex behavior and responses required from robots operating in urban areas and human living spaces are handled through prioritized, non-overlapping tasks defined using null-spaces. Accurate measurement of contact forces is crucial for robotic applications that involve interaction with the surrounding environment. However, measurements obtained from force sensors are generally noisy and need to be processed before they can be used as control inputs. Furthermore, the use of force/torque sensors can be impracticable or even impossible in some applications. We are investigating robust force estimation methods that can augment force/torque sensors or all together replace them. In conjunction, we are conducting research on motion-force control of underactuated robots.
Teleoperation and Haptics
The past two decades has seen the tremendous emergence of robotic surgery. Today, 90% of some medical procedures are performed with surgical robots. This trend is expected to continue with an increasing pace in the coming years. We are currently developing robotic systems for minimally invasive surgery in collaboration with Marmara University as a part of a TUBITAK funded research project (1003 program call SB0206, project no: 15E712, 2016-2018). Bilateral haptic teleoperation, advanced force-motion control, and supervisory control schemes are currently being investigated by our team. We are also conducting research on model-mediated teleoperation based on multi-modal perception to alleviate the adverse affects of time delays and data loss on the stability of teleoperation methods.
Perception and Autonomous Navigation
The lab conducts research on multi-modal perception based on novel sensing modalities. Path planning, autonomous navigation, and adaptation to dynamically updated maps are the main lines of research that fall under this research agenda.
The conventional approach for modeling, dynamic analysis, and control of robotic systems is based on the assumption that the robots are rigid. Relatively small inherent flexibility is ignored in the modeling stage and their adverse effects are compensated by feedback controllers. Soft robotics is an emerging field that focuses on the theory, design, control, manufacturing, and applications of robots built using soft and highly deformable elastic materials such as silicone rubber. These soft robots offer significant improvements in safety, manipulability, cost, and ease of manufacturing (using 3D printing techniques). In the past several years, soft robotics research has resulted in a variety of successful applications that involve contact and interact of the robot with its surroundings. These applications include mobile robots for rough and unstructured terrain, underwater robotic arms, and more recently, robotic grippers Despite this rapidly expanding literature, soft robots still pose several major challenges that reduce their dexterity and thus, their widespread utilization. Soft and very deformable materials exhibit significant levels of structural non-linearity and are difficult to analytically model. As a consequence, shape, motion, and force control of soft robots is very challenging. We are currently working on developing structural model, advanced control architectures, and manufacturing techniques of soft mobile robots and soft underwater robotic arms.
Robot Learning and Human-Robot Interaction
Artificial intelligence and machine learning methods are employed for teaching robots skills through human demonstrations and previous experiences.