Robotics Science is the scientific and engineering study of robots and robotic systems, including their mechanisms, sensing, computation, control, planning, interaction, learning, autonomy, and behaviour. A robot is generally a physical system capable of sensing aspects of its environment and producing controlled physical actions, although the exact boundary of the term varies across contexts. Robotics therefore combines mechanical engineering, electrical engineering, computer science, control theory, artificial intelligence, materials science, mathematics, human factors, and increasingly neuroscience and cognitive science.
The scientific questions of robotics include how machines can perceive the world, estimate their state, plan safe and efficient actions, manipulate objects, navigate environments, interact with people, and adapt when conditions differ from expectations. Robotics Science is distinct from the study of a particular robot because it seeks general principles and methods that can be applied across robotic platforms and tasks.
The history of robotics includes ancient mechanical automata, industrial mechanisation, teleoperation, programmable manipulators, mobile robots, and increasingly autonomous systems. The word robot derives from the Czech term robota, associated with forced labour or compulsory work, and entered international usage through Karel Capek's 1920 play R U R. Industrial robots became prominent during the twentieth century, particularly in manufacturing tasks involving repetitive and programmable manipulation.
Early robots were often highly structured machines operating in controlled environments, where positions and tasks could be specified precisely. Research later expanded into mobile robotics, autonomous navigation, manipulation in unstructured environments, humanoid systems, medical robotics, aerial robotics, underwater robotics, agricultural robots, and collaborative robots. Modern robotics increasingly integrates machine learning and advanced sensing, but many dependable robotic systems still rely on carefully engineered control and planning methods because physical safety and predictable behaviour are important.
Robot morphology describes the physical structure through which a robot interacts with its environment. A manipulator can contain joints and links arranged in configurations such as serial or parallel mechanisms, while mobile robots can use wheels, tracks, legs, propellers, or other locomotion systems. Degrees of freedom describe independent motions available to a mechanism, and kinematics relates joint configurations to the position and orientation of the robot's end effector or body.
Forward kinematics calculates the resulting pose from known joint states, while inverse kinematics seeks joint configurations capable of producing a desired pose. Dynamics concerns the relationship between forces, torques, masses, accelerations, and motion. Mechanical design must account for stiffness, weight, energy consumption, actuator capabilities, friction, backlash, thermal constraints, and physical durability. These mechanical properties directly influence what forms of control and planning are feasible.
Robotic perception depends on sensors that provide information about the robot and its environment. Common sensors include cameras, depth cameras, lidar, radar, inertial measurement units, encoders, force sensors, tactile sensors, microphones, and proximity sensors. Sensor fusion combines information from multiple sources to obtain more reliable estimates than may be possible from one sensor alone. State estimation determines quantities such as robot position, velocity, orientation, and sometimes the state of environmental objects.
Localisation determines where a robot is relative to a map or reference frame, while mapping constructs a representation of the environment. Simultaneous localisation and mapping addresses the coupled problem in which the robot must estimate both its own position and an initially uncertain map. Perception remains difficult because sensors are noisy, environments can change, objects can be occluded, lighting can vary, and real world measurements rarely correspond perfectly to ideal mathematical assumptions.
Planning and control translate goals into physical actions. Motion planning searches for feasible trajectories that respect robot geometry, obstacles, kinematic constraints, dynamic limits, and task requirements. Sampling based methods, optimisation based methods, graph search, and reactive techniques provide different approaches to this problem. Control systems determine actuator commands required to follow desired states or trajectories. Feedback control uses measurements of the current state to adjust actions, helping compensate for disturbances and modelling errors.
Proportional integral derivative control remains widely used because of its simplicity and effectiveness, while more advanced methods include model predictive control, adaptive control, robust control, impedance control, and nonlinear control. A robot's planner and controller must be coordinated because a mathematically valid trajectory may be physically infeasible or difficult to execute accurately.
Robot learning applies Machine Learning methods to perception, control, planning, manipulation, and adaptation. Supervised learning can map sensor observations to classifications or predicted states, reinforcement learning can learn policies through interaction, and imitation learning can learn behaviours from demonstrations. Learning based approaches can reduce the need for manually specifying complex behaviours, but physical robots introduce challenges that are less severe in purely digital environments. Data collection can be expensive, experiments can damage hardware, and unsafe actions can create hazards. Simulation can provide large amounts of training data, but the difference between simulated and physical environments creates the problem commonly called the reality gap. Domain randomisation, system identification, simulation to real transfer, and real world fine tuning are methods used to address this difference.
Human robot interaction studies how people and robots communicate, collaborate, and share physical or cognitive tasks. Robots operating near people need to account for human motion, expectations, attention, uncertainty, and safety. Collaborative robots can be designed to work in shared spaces rather than behind physical barriers, which creates additional requirements for sensing, force limitation, collision detection, and safe control.
Social robots may use speech, gestures, facial displays, or other signals, but apparent social intelligence should not automatically be equated with human understanding. Medical robots, rehabilitation robots, assistive robots, warehouse systems, autonomous vehicles, drones, and industrial manipulators each have distinct safety and reliability requirements. The appropriate level of autonomy depends on the consequences of failure and the ability of humans to monitor and intervene.
Robotics Science faces fundamental challenges because physical environments are uncertain, dynamic, partially observable, and governed by mechanical constraints. A robot may have to act despite incomplete information while avoiding collisions, conserving energy, coping with unexpected objects, and maintaining reliable operation. Robustness is therefore often more important than performance on an ideal benchmark. Other limitations include actuator wear, sensor failure, computational latency, communication delays, limited battery capacity, environmental interference, maintenance requirements, and difficulties generalising learned behaviours.
The field is increasingly concerned with embodied intelligence, in which perception, action, morphology, and environmental interaction are considered together rather than as independent modules. Robotics demonstrates that intelligence in physical systems cannot be reduced entirely to abstract computation because actions have mechanical consequences, observations are constrained by sensors, and successful behaviour depends on continuous interaction between a machine and its environment.
