Optimizing the body and brain of a robot is a coupled challenge: the morphology determines what control strategies are effective, while the control parameters influence how well the morphology performs. This joint optimization can be done through nested loops of evolutionary and learning processes, where the control parameters of each robot are learned independently. However, the control parameters learned by one robot may contain valuable information for others. Thus, we introduce a social learning approach in which robots can exploit optimized parameters from their peers to accelerate their own brain optimization. Within this framework, we systematically investigate how the selection of teachers, deciding which and how many robots to learn from, affects performance, experimenting with virtual soft robots in four tasks and environments. In particular, we study the effect of inheriting experience from morphologically similar robots due to the tightly coupled body and brain in robot optimization. Our results confirm the effectiveness of building on others' experience, as social learning clearly outperforms learning from scratch under equivalent computational budgets. In addition, while the optimal teacher selection strategy remains open, our findings suggest that incorporating knowledge from multiple teachers can yield more consistent and robust improvements.

The brain's distributed architecture has inspired numerous artificial intelligence (AI) systems, particularly through its neocortical organization. However, current AI approaches largely overlook a crucial aspect of biological intelligence: active sensing - the deliberate movement of sensory organs to explore the environment. To explore how sensor movement impacts behavior in image classification tasks, we introduce the Active Neural Cellular Automata (ANCA), a neocortex-inspired model with movable sensors. Active sensing naturally emerges in the ANCA, with belief-informed exploration and attentive behavior to salient information, without adding explicit attention mechanisms. We show that active sensing simplifies classification tasks. Moreover, active sensing lets the ANCA be smaller than the image size without losing information, which makes it highly scalable. We show that the ANCA maintains over 90% accuracy zero-shot as the system size is increased or decreased on a 3-class MNIST task. This scalability enables fault tolerance on the same task, maintaining over 90% accuracy with up to 70% silenced sensors, a scenario where traditional architectures fail. Overall, our work provides insight to how distributed architectures can interact with movement, opening new avenues for adaptive AI systems in embodied agents.

Entrainment of movement to a periodic stimulus is a characteristic intelligent behaviour in humans and an important goal for adaptive robotics. We demonstrate a quadruped central pattern generator (CPG), consisting of modified Matsuoka neurons, that spontaneously adjusts its period of oscillation to that of a periodic input signal. This is done by simple forcing, with the aid of a filtering network as well as a neural model with tonic input-dependent oscillation period. We first use the NSGA3 algorithm to evolve the CPG parameters, using separate fitness functions for period tunability, limb homogeneity and gait stability. Four CPGs, maximizing different weighted averages of the fitness functions, are then selected from the Pareto front and each is used as a basis for optimizing a filter network. Different numbers of neurons are tested for each filter network. We find that period tunability in particular facilitates robust entrainment, that bounding gaits entrain more easily than walking gaits, and that more neurons in the filter network are beneficial for pre-processing input signals. The system that we present can be used in conjunction with sensory feedback to allow low-level adaptive and robust behaviour in walking robots.

Generalizability is seen as one of the major challenges in deep learning, in particular in the domain of medical imaging, where a change of hospital or in imaging routines can lead to a complete failure of a model. To tackle this, we introduce Consistency Training, a training procedure and alternative to data augmentation based on maximizing models’ prediction consistency across augmented and unaugmented data in order to facilitate better out-of-distribution generalization. To this end, we develop a novel auxiliary region-based segmentation loss function called Segmentation Inconsistency Loss (SIL), which considers the differences between pairs of augmented and unaugmented predictions and labels. We demonstrate that Consistency Training outperforms conventional data augmentation on several out-of-distribution datasets on polyp segmentation, a popular medical task.

Robots operating in the real world will experience a range of different environments and tasks. It is essential for the robot to have the ability to adapt to its surroundings to work efficiently in changing conditions. Evolutionary robotics aims to solve this by optimizing both the control and body (morphology) of a robot, allowing adaptation to internal, as well as external factors. Most work in this field has been done in physics simulators, which are relatively simple and not able to replicate the richness of interactions found in the real world. Solutions that rely on the complex interplay among control, body, and environment are therefore rarely found. In this article, we rely solely on real-world evaluations and apply evolutionary search to yield combinations of morphology and control for our mechanically self-reconfiguring quadruped robot. We evolve solutions on two distinct physical surfaces and analyze the results in terms of both control and morphology. We then transition to two previously unseen surfaces to demonstrate the generality of our method. We find that the evolutionary search finds high-performing and diverse morphology-controller configurations by adapting both control and body to the different properties of the physical environments. We additionally find that morphology and control vary with statistical significance between the environments. Moreover, we observe that our method allows for morphology and control parameters to transfer to previously unseen terrains, demonstrating the generality of our approach.

Neural networks have been widely used in agent learning architectures; however, learnings for one task might nullify learnings for another. Behavioural plasticity enables humans and animals alike to respond to environmental changes without degrading learned knowledge; this can be achieved by regulating behaviour with neuromodulation—a biological process found in the brain. We demonstrate that by modulating activity-propagating signals, neurally trained agents evolving to solve tasks in dynamic environments that are prone to change can expect a significantly higher fitness than non-modulatory agents and also achieve their goals more often. Further, we show that while behavioural plasticity can help agents to achieve goals in these variable environments, this ability to overcome environmental changes with greater success comes at the cost of highly volatile evolution.

Overcoming robotics challenges in the real world requires resilient control systems capable of handling a multitude of environments and unforeseen events. Evolutionary optimization using simulations is a promising way to automatically design such control systems, however, if the disparity between simulation and the real world becomes too large, the optimization process may result in dysfunctional real-world behaviors. In this paper, we address this challenge by considering embodied phase coordination in the evolutionary optimization of a quadruped robot controller based on central pattern generators. With this method, leg phases, and indirectly also inter-leg coordination, are influenced by sensor feedback. By comparing two very similar control systems we gain insight into how the sensory feedback approach affects the evolved parameters of the control system, and how the performances differ in simulation, in transferal to the real world, and to different real-world environments. We show that evolution enables the design of a control system with embodied phase coordination which is more complex than previously seen approaches, and that this system is capable of controlling a real-world multi-jointed quadruped robot. The approach reduces the performance discrepancy between simulation and the real world, and displays robustness towards new environments.

Collaborative robots are becoming more common on factory floors as well as regular environments, however, their safety still is not a fully solved issue. Collision detection does not always perform as expected and collision avoidance is still an active research area. Collision avoidance works well for fixed robot-camera setups, however, if they are shifted around, Eye-to-Hand calibration becomes invalid making it difficult to accurately run many of the existing collision avoidance algorithms. We approach the problem by presenting a stand-alone system capable of detecting the robot and estimating its position, including individual joints, by using a simple 2D colour image as an input, where no Eye-to-Hand calibration is needed. As an extension of previous work, a two-stage transfer learning approach is used to re-train a multi-objective convolutional neural network (CNN) to allow it to be used with heterogeneous robot arms. Our method is capable of detecting the robot in real-time and new robot types can be added by having significantly smaller training datasets compared to the requirements of a fully trained network. We present data collection approach, the structure of the multi-objective CNN, the two-stage transfer learning training and test results by using real robots from Universal Robots, Kuka, and Franka Emika. Eventually, we analyse possible application areas of our method together with the possible improvements.

If robots are to become ubiquitous, they will need to be able to adapt to complex and dynamic environments. Robots that can adapt their bodies while deployed might be flexible and robust enough to meet this challenge. Previous work on dynamic robot morphology has focused on simulation, combining simple modules, or switching between locomotion modes. Here, we present an alternative approach: a self-reconfigurable morphology that allows a single four-legged robot to actively adapt the length of its legs to different environments. We report the design of our robot, as well as the results of a study that verifies the performance impact of self-reconfiguration. This study compares three different control and morphology pairs under different levels of servo supply voltage in the lab. We also performed preliminary tests in different uncontrolled outdoor environments to see if changes to the external environment supports our findings in the lab. Our results show better performance with an adaptable body, lending evidence to the value of self-reconfiguration for quadruped robots.

In recent years, researchers have explored human body posture and motion to control robots in more natural ways. These interfaces require the ability to track the body movements of the user in three dimensions. Deploying motion capture systems for tracking tends to be costly and intrusive and requires a clear line of sight, making them ill adapted for applications that need fast deployment. In this article, we use consumer-grade armbands, capturing orientation information and muscle activity, to interact with a robotic system through a state machine controlled by a body motion classifier. To compensate for the low quality of the information of these sensors, and to allow a wider range of dynamic control, our approach relies on machine learning. We train our classifier directly on the user to recognize (within minutes) which physiological state his or her body motion expresses. We demonstrate that on top of guaranteeing faster field deployment, our algorithm performs better than all comparable algorithms, and we detail its configuration and the most significant features extracted. As the use of large groups of robots is growing, we postulate that their interaction with humans can be eased by our approach. We identified the key factors to stimulate engagement using our system on 27 participants, each creating his or her own set of expressive motions to control a swarm of desk robots. The resulting unique dataset is available online together with the classifier and the robot control scripts.

For robots to handle the numerous factors that can affect them in the real world, they must adapt to changes and unexpected events. Evolutionary robotics tries to solve some of these issues by automatically optimizing a robot for a specific environment. Most of the research in this field, however, uses simplified representations of the robotic system in software simulations. The large gap between performance in simulation and the real world makes it challenging to transfer the resulting robots to the real world. In this paper, we apply real world multi-objective evolutionary optimization to optimize both control and morphology of a four-legged mammal-inspired robot. We change the supply voltage of the system, reducing the available torque and speed of all joints, and study how this affects both the fitness, as well as the morphology and control of the solutions. In addition to demonstrating that this real-world evolutionary scheme for morphology and control is indeed feasible with relatively few evaluations, we show that evolution under the different hardware limitations results in comparable performance for low and moderate speeds, and that the search achieves this by adapting both the control and the morphology of the robot.

Percutaneous needle insertion guided by ultrasound imaging is routinely performed in hospitals today. Automating these procedures could increase placement accuracy and lower time usage of health care personnel to perform these procedures. An important step in the automation is the estimation of the needle orientation and position in the ultrasound image. One approach to estimate the needle orientation and position is to have the needle aligned with the image plane of the ultrasound probe. Aligning the needle with the plane is difficult, even with accurate measurements and calibration of both needle and probe. In this paper we propose a visual servoing method to move the ultrasound probe, using a robot to align the image plane of the probe with the needle. The method segments the needle and updates a set of visual features based on a model of the needle. A state machine is used to keep track of the alignment process, and different visual features are used to control the probe in the different states. Both simulation using simple synthetic images and experiments in a water tank are conducted to validate the proposed method. The simulation shows that the proposed method manages to align the probe plane with the needle. The alignment process is slower in the actual robot experiment, but still aligns the probe plane with the needle. The method is shown to work under simplified conditions, and is a step towards a method that could be applied in a clinical setting.

The evolvable hardware (EHW) paradigm relies on continuous run-time reconfiguration of hardware. When applied on modern FPGAs, the technically challenging reconfiguration process becomes an issue and can be approached at multiple levels. In related work, virtual reconfigurable circuits (VRC), partial reconfiguration, and lookup table (LUT) reconfiguration approaches have been investigated. In this paper, we show how fine-grained partial reconfiguration of 6-input LUTs of modern Xilinx FPGAs can lead to significantly more efficient resource utilization in an EHW application. Neither manual placement nor any proprietary bitstream manipulation is required in the simplest form of the employed method. We specify the goal architecture in VHDL and read out the locations of the automatically placed LUTs for use in an online reconfiguration setting. This allows for an easy and flexible architecture specification, as well as possible implementation improvements over a hand-placed design. For demonstration, we rely on a hardware signal classifier application. Our results show that the proposed approach can fit a classification circuit 4 times larger than an equivalent VRC-based approach, and 6 times larger than a shift register-based approach, in a Xilinx Virtex-5 device. To verify the reconfiguration process, a MicroBlaze-based embedded system is implemented, and reconfiguration is carried out via the Xilinx Internal Configuration Access Port (ICAP) and driver software.

Evolutionary robotics often involves optimization in large, complex search spaces, requiring good population diversity. Recently, measures to actively increase diversity or novelty have been employed in order to get sufficient exploration of the search space either as the sole optimization objective or in combination with some performance measurement. When evolving morphology in addition to the control system, it can be difficult to construct a measure that sufficiently captures the qualitative differences between individuals. In this paper we investigate four diversity measures, applied in a set of evolutionary robotics experiments using an indirect encoding for evolving robot morphology. In the experiments we optimize forward locomotion capabilities of symmetrical legged robots in a physics simulation.

This paper proposes an approach to representing robot morphology and control, using a two-level description linked to two different physical axes of development. The bioinspired encoding produces robots with animal-like bilateral limbed morphology with co-evolved control parameters using a central pattern generator-based modular artificial neural network. Experiments are performed on optimizing a simple simulated locomotion problem, using multi-objective evolution with two secondary objectives. The results show that the representation is capable of producing a variety of viable designs even with a relatively restricted set of parameters and a very simple control system. Furthermore, the utility of a cumulative encoding over a non-cumulative approach is demonstrated. We also show that the representation is viable for real-life reproduction by automatically generating CAD files, 3D printing the limbs, and attaching off-the-shelf servomotors.

Evolvable hardware (EHW) has shown itself to be a promising approach for prosthetic hand controllers. Besides competitive classification performance, EHW classifiers offer self-adaptation, fast training, and a compact implementation. However, EHW classifiers have not yet been sufficiently compared to state-of-the-art conventional classifiers. In this paper, we compare two EHW approaches to four conventional classification techniques:k-nearest-neighbor, decision trees, artificial neural networks, and support vector machines. We provide all classifiers with features extracted from electromyographic signals taken from forearm muscle contractions, and let the algorithms recognize eight to eleven different kinds of hand movements. We investigate classification accuracy on a fixed data set and stability of classification error rates when new data is introduced. For this purpose, we have recorded a short-term data set from three individuals over three consecutive days and a long-term data set from a single individual over three weeks. Experimental results demonstrate that EHW approaches are indeed able to compete with state-of-the-art classifiers in terms of classification performance.

A lot of optimal and heuristic algorithms for solving facility layout problem (FLP) have been developed in the past few decades. The majority of these approaches adopt a problem formulation known as the quadratic assignment problem (QAP) that is particularly suitable for equal area facilities. Unequal area FLP comprises a class of extremely difficult and widely applicable optimization problems arising in many diverse areas to meet the requirements for real-world applications. Unfortunately, most of these approaches are based on a single objective. While, the real-world FLPs are multi-objective by nature. Only very recently have meta-heuristics been designed and used in multi-objective FLP. They most often use the weighted sum method to combine the different objectives and thus, inherit the well-known problems of this method. As of now, there is no formal approach published for the unequal area multi-objective FLP to consider several objectives simultaneously. This paper presents an evolutionary approach for solving multi-objective unequal area FLP using multi-objective genetic algorithm that presents the layout as a set of Pareto-optimal solutions optimizing multiple objectives simultaneously. The experimental results show that the proposed approach performs well in dealing with multi-objective unequal area FLPs which better reflects the real-world scenario.

A Coevolutionary, Hyper Heuristic approach to the optimization of Three-dimensional Process Plant Layouts (3DPPLs) is explored. By taking advantage of the natural problem decomposition, one population of layout heuristics, and another population of scheduling heuristics are coevolved. Generalized heuristics are evolved by training on multiple small problem instances, so that training time is reduced. The best generalized heuristic builds arbitrary sized 3DPPLs which reduce the cost by 18% when compared to a handmade heuristic. Specialized heuristics are evolved by optimizing each problem instance and outperforms the generalized heuristics after a fixed number of generations. Compared to a direct-encoded Genetic Algorithm, the benefit of specialized heuristics increases with the size of the problem, and costs are reduced by 30% when compared to the handmade heuristic.

Type-2 fuzzy logic systems are proposed as an alternative solution in the literature when a system has a large amount of uncertainties and type-1 fuzzy systems come to the limits of their performances. In this study, an adaptive type-2 fuzzy-neuro system is designed for the position control of a servo system with an intelligent sensor. The sensor gives different resistance values with respect to the stretch of it, and it is supposed to be used in an robotic arm position measurement system. These kinds of sensors can be used in human-assistance robots that have soft surfaces in order not to damage the humans. However, these sensors have time-varying gains and uncertainties that are not very easy to handle. Moreover, they generally have a hysteresis on their input-output relations. The simulation results show that the control algorithm developed gives better performances when compared to conventional type-1 fuzzy controllers on such a highly nonlinear, uncertain system.

This paper investigates advanced features of the PhysX physics simulation engine for simulated robotic evolution, with the goal of applying the results to a real world soft robotic system which is under construction. The cloth feature in PhysX has the potential of taking into account complex dynamics while at the same time being accelerated by a graphics processing unit. As an initial approach, muscle-shaped structures are simulated with the cloth feature and employed as actuators in a robotic structure where both morphology and control parameters are subject to optimization by a genetic algorithm. A linear and a spring-damper-based model have also been applied for reference. Stable locomotion has been successfully evolved, however, attention to simulation parameters has been necessary in order to avoid simulator instability.

Efficient heuristics are required for on-line optimization problems where search-based methods are unfeasible due to frequent dynamics in the environment. This is especially apparent when operating on combinatorial NP-complete problems involving a large number of items. However, designing new heuristics for these problems may be a difficult and time-consuming task even for domain experts. Therefore, automating this design process may benefit the industry when facing new and difficult optimization problems. The Distributor's Pallet Packing Problem (DPPP) is the problem of loading a pallet of non-homogenous items coming off a production line and is an instance of a range of resource-constrained, NP-complete, scheduling problems that are highly relevant for practical tasks in the industry. Common heuristics for the DPPP typically decompose the problem into two sub-problems; one of pre-scheduling all items on the production line and one of packing the items on the pallet. In this paper we concentrate on a two dimensional version of the DPPP and the more realistic scenario of having knowledge about only a limited set of the items on the production line. This paper aims at demonstrating that such an unknown heuristic may be evolved by Gene Expression Programming and Cooperative Coevolution. By taking advantage of the natural problem decomposition, two species evolve heuristics for pre-scheduling and packing respectively. We also argue that the evolved heuristics form part of a developmental stage in the construction of the finished phenotype, that is, the loaded pallet.

Several practical problems in industry are difficult to optimize, both in terms of scalability and representation. Heuristics designed by domain experts are frequently applied to such problems. However, designing optimized heuristics can be a non-trivial task. One such difficult problem is the Facility Layout Problem (FLP) which is concerned with the allocation of activities to space. This paper is concerned with the block layout problem, where the activities require a fixed size and shape (modules). This problem is commonly divided into two sub problems; one of creating an initial feasible layout and one of improving the layout by interchanging the location of activities. We investigate how to extract novel heuristics for the FLP by applying an approach called Cooperative Coevolutionary Gene Expression Programming (CCGEP). By taking advantage of the natural problem decomposition, one species evolves heuristics for pre-scheduling, and another for allocating the activities onto the plant. An experimental, comparative approach investigates various features of the CCGEP approach. The results show that the evolved heuristics converge to suboptimal solutions as the problem size grows. However, coevolution has a positive effect on optimization of single problem instances. Expensive fitness evaluations may be limited by evolving generalized heuristics applicable to unseen fitness cases of arbitrary sizes.

A pattern recognition system that can process a large amount of image data at high speed is required in many fields. In this paper, we propose an on-chip pattern recognition system that utilizes the reconfigurability of the FPGA. The features of the system are not only very high recognition speed but also an adaptive function. For example, when objects to be detected change appearance, recognition parameters must be changed to retain the recognition accuracy. The system can automatically adjust by executing on-chip partial reconfiguration. The system runs at 25MHz and can return a recognition result in one clock cycle, 40ns. To update the system, all processes needed for searching for the best recognition parameters, generating configuration data and reconfiguring the system are carried out within 30s.