Intelligent Autonomous Systems 16: Proceedings of the 16th International Conference IAS-16 (Lecture Notes in Networks and Systems, 412)

This book presents the latest advances and research achievements in the fields of autonomous robots and intelligent systems, presented at the IAS-16 conference, conducted virtually in Singapore, from 22 to 25 June 2021. IAS is a common platform for an exchange and sharing of ideas among the international scientific research and technical community on some of the main trends of robotics and autonomous systems: navigation, machine learning, computer vision, control, and robot design―as well as a wide range of applications. IAS-16 reflects the rise of machine learning and deep learning developments in the robotics field, as employed in a variety of applications and systems. All contributions were selected using a rigorous peer-reviewed process to ensure their scientific quality. Despite the challenge of organising a conference during a pandemic, the IAS biennial conference remains an essential venue for the robotics and autonomous systems community ever since its inception in 1986. Chapters 46 of this book is available open access under a CC BY 4.0 license at link.springer.com Preface Organization Program Committee Contents Localization and SLAM Online Learning Based Long-Term Feature Existence State Prediction for Visual Topological Localization 1 Introduction 2 Online Visual Topological Localization System 2.1 Image Pre-processing 2.2 Feature Matching 2.3 Feature Existence Matrix Maintenance 2.4 Online Time Series Modeling 3 Online Constrained Autoregressive Newton Step Method 3.1 Time Series Model and Preliminaries 3.2 Autoregressive Order Determination and Parameter Fitting 3.3 Online Learning for Model Parameter Updating 4 Simulation and Experimental Results 4.1 Modeling Simulated Varying Periodic Time Series 4.2 Performance on Real Measured Time Series Datasets 4.3 Topological Localization Accuracy 4.4 Real-Time Performance Analysis 5 Conclusion and Future Work References ALOT: Augmented Localization with Obstacle Tracking 1 Introduction 2 Background and Related Work 3 System Architecture 3.1 Fusion of 3D LiDAR Scan Clustering with Vision Classification 3.2 Dynamic Multi-object Tracking 3.3 Ego-Pose Proposals 3.4 Dynamic Obstacle Removal from Flattened 3D LiDAR Scan 3.5 Particle Filter Localization with Dual Particle Weighting 4 Experimental Results and Discussion 4.1 Experimental Hardware and Set-Up 4.2 Data Collection and Bench-Marking 4.3 Localization Results Under Light to Moderate Occlusion 4.4 Localization Results Under Heavy Occlusion 5 Conclusion References RTS-vSLAM: Real-Time Visual Semantic Tracking and Mapping Under Dynamic Environments 1 Introduction 2 Related Work 2.1 Geometric-Based Methods 2.2 Semantic-Based Methods 3 System Overview 4 Server 4.1 Semantic Generator 4.2 Mapping 5 Client 5.1 Feature Cluster Algorithm 5.2 Tracking Algorithm 5.3 Geometry Consistency Check Algorithm 6 Experimental Results 6.1 Evaluation on TMU Dataset 6.2 Result of Mapping 6.3 Computation Time Analysis 7 Conclusion References Mobile Robot Localization Considering the Attributes of Objects to Prevent the Kidnapped Robot Problem 1 Introduction 2 Semantic Localization Considering Attributes of Objects 2.1 Object Class and Fixed-Rate in Localization 2.2 Proposed Model 3 Simulation 3.1 Method Used for Comparison 3.2 Definition of the Kidnapped Robot Problem 3.3 Simulation Scenario and Conditions 3.4 Simulation Results 4 Experiment 4.1 Experimental Platform 4.2 Pre-validation 4.3 Experiment Results 5 Conclusion References Collaborative Mapping of Archaeological Sites Using Multiple UAVs 1 Introduction 2 Related Work 2.1 Archaeological Mapping Using UAVs 2.2 Collaborative SLAM 3 Methodology 3.1 CCM-SLAM Architecture 3.2 The Agents 3.3 The Server 4 Environment Setup and Dataset 4.1 CR AirSim Environment 4.2 CR Monocular Cam Dataset 4.3 Real-World Setup and the Sadra Fort (SF) Dataset 5 Experiments and Evaluation 5.1 Single-Agent Evaluation on CR Dataset 5.2 Collaborative SLAM with Significant Trajectory Overlap 5.3 Collaborative Mapping of CR Environment 5.4 Collaborative Mapping of Sadra Fort 5.5 Dense Reconstruction of Sadra Fort 6 Conclusion and Future Work References 3D Nominal Scene Reconstruction for Object Localization and UAS Navigation 1 Introduction 2 Literature Review 3 Initial Study 4 Baseline Meshing with a Depth Sensing Camera 5 Photogrammetry Suite: Agisoft Metashape 5.1 Study Room Environment 5.2 Forested Area Environment 6 Conclusion References Navigation Topometric Navigation Considering Movable Objects 1 Introduction 2 Environment Recognition 2.1 Object Recognition and Position Estimation 2.2 Movability of Object Estimation 3 Topometric Navigation 3.1 Generating Topological Map 3.2 Generating Metric Map 3.3 Path Planning 4 Experiment 4.1 Experiment Conditions 4.2 System Configuration 4.3 Experiment Results 5 Conclusion References Harmonious Robot Navigation Strategies for Pedestrians 1 Introduction 2 Social Force Model (SFM) 3 Design of Motion Control to Reduce the Impact on Surrounding Pedestrians 3.1 Determination of the Robot Velocity Command 3.2 Design of Evaluation Values 3.3 Design of Weight i ,i 4 Simulation 4.1 Simulation Conditions 4.2 Simulation Result 5 Conclusion References Shared Autonomy for Telepresence Robots Based on People-Aware Navigation 1 Introduction 2 Shared People-Aware Navigation 3 Telepresence Interface 3.1 Experiment 3.2 Participants 3.3 Protocol 3.4 Robotic Platform 3.5 Navigation System 3.6 People Detection 3.7 Questionnaire 4 Results 4.1 Shared vs. Manual and Shared vs. Supervisory Modalities 4.2 Assistance in Shared Modality 4.3 Human Evaluation 5 Conclusion References A Framework to Study Autonomous Driving User Acceptance in the Wild 1 Introduction 2 Related Work 3 Physiological Data Collection and Feature Extraction 3.1 Physiological Data Acquired 3.2 External Conditions Understanding 4 Methods for Data Analysis 4.1 Event Sequence Identification 4.2 Correlation Analysis 5 Acquisition Campaign 5.1 Participants and Experimental Setup 5.2 Protocol 5.3 Experimental Results 5.4 Questionnaire Results 5.5 Discussion 6 Conclusion and Future Work References Probabilistic Collision Constraint for Motion Planning in Dynamic Environments 1 Introduction 2 Related Work 3 Preliminaries 4 Probabilistic Collision Avoidance 4.1 Collision Constraint 4.2 Objective Function 4.3 Comparison to Other Approaches 5 Results 6 Conclusion References Human-Following Control in Furrow For Agricultural Support Robot 1 Introduction 2 Related Works 3 System Overview 4 Ridge Detection and Interpolation 4.1 Extraction of Ridge Line Candidates from Image and PCD 4.2 Ridge Line Tracking and Interpolation 5 User-Following Control in Furrows 5.1 MF for User-Following 5.2 MF for Ridge Avoidance 5.3 Velocity Command Based on Mixed MF 6 Experiments 7 Conclusion References Autonomous Off-Road Navigation Using Near-Feature-Based World Knowledge Incorporation on the Example of Forest Path Detection 1 Introduction 2 Related Work 3 Behavior-Based Control 4 Forest Way Detection 5 Local Multi-Feature Map 5.1 Ground Features 5.2 Obstacle Features 6 Forest-Path-based Localization 6.1 OpenStreepMap Data 6.2 Pathway Matching for Local Localization 7 Navigation 7.1 Local Navigation 7.2 Behavior-Based Realization 8 Experiments 8.1 Forest Way Detection 8.2 Multi-feature Navigation 9 Conclusion References Learning to Drive Fast on a DuckieTown Highway 1 Introduction 2 Related Work 3 Method 3.1 Reinforcement Learning Framework 3.2 Network Architecture 3.3 Reward Function and Setup 4 Results 4.1 Simulator Evaluation 4.2 Real-World Evaluation 5 Discussion 6 Conclusion References End-to-End Learning of Autonomous Vehicle Lateral Control via MPC Training 1 Introduction 2 Related Works 3 Dataset Generation 3.1 Data Acquisition 3.2 Ground Truth 4 System Architecture and Training 5 Experimental Results 5.1 Offline Results 5.2 In-the-Loop Validation 6 Conclusion and Future Works References End-to-End Imitation Learning for Autonomous Vehicle Steering on a Single-Camera Stream 1 Introduction 2 Related Work 3 Experiment 4 Result 5 Discussion 6 Conclusion References Application Sensorized Compliant Robot Gripper for Estimating the Cooking Time of Boil-Cooked Vegetables 1 Introduction 2 Methods 2.1 Vegetable Mechanical Model 2.2 Kinematic Model of Vegetable 2.3 Fitting the Model 3 Hardware Development 3.1 Gripper Technology 3.2 Sensing Technologies 3.3 Stiffness Estimation 4 Results 4.1 Stiffness Measurement 4.2 Model Computing 4.3 Model Predictions 5 Discussion and Conclusions References A Study on Constraint-Free Vital Sign Measurement Using a Millimeter Wave Sensor in Home Environment 1 Introduction 1.1 The Need for Home Monitoring 1.2 Previous Research 1.3 Research Objectives 2 Proposed Method 3 Experiment 4 Results 5 Discussion 6 Conclusion References Development of a Medication-Taking Behavior Monitoring System Using Depth Sensor 1 Introduction 2 Methods 2.1 Medication Adherence Monitoring System 2.2 Experiments 3 Results and Discussion 3.1 Comparison Of Grasp Identification Methods 3.2 Completion Detection Of Medication-Taking Action 3.3 Interruption Detection of Medication-Taking Action 4 Conclusion References Developing a Collaborative Robotic Dishwasher Cell System for Restaurants 1 Introduction 2 Task and Rate Requirement Capture 2.1 Task Sequence 2.2 Rate Requirement 3 System Design 3.1 Description of the System 3.2 Task Controller and Path Planner Component 3.3 Image Recognition Component 3.4 Actuated Suction Cup End-Effector Component 3.5 Dish Loading Planner Component 3.6 Dish Unloading Planner Component 4 Design Validation 5 Conclusions References Simulated Annealing Algorithms for the Heterogeneous Robots Task Scheduling Problem in Heterogeneous Robotic Order Fulfillment Systems 1 Introduction 2 Problem Statement 3 Iterative Improvement by Simulated Annealing 3.1 Simulated Annealing Basics 3.2 Solution Representation and Objective Function 3.3 Initial Solution Generation 3.4 Neighborhood Structure 3.5 Annealing and Cooling Schemes 4 Computation Results and Analysis 4.1 Dataset Design 4.2 Results and Analysis 5 Conclusions and Future Work References Robotic Cooking Through Pose Extraction from Human Natural Cooking Using OpenPose 1 Introduction 2 Materials and Methods 2.1 Recording of Human Cooking 2.2 Trajectory Analysis with OpenPose 2.3 Robotic Chef 3 Experiments 3.1 Human Demonstrations 3.2 Evaluation Protocol 4 Results and Discussion 5 Conclusion and Future Work References Multiple Object Detection and Segmentation for Automated Removal in Additive Manufacturing with Service Robots 1 Introduction 2 Approach 2.1 Determination of the Object Hypothesis Based on G-code 2.2 Verification of the Object Hypothesis Based on a Point Cloud 3 Experiments and Results 4 Conclusion and Outlook References Determination of Posture Comfort Zones for Robot-Human Handover Tasks 1 Introduction 2 Concept 3 Experiments 3.1 Experimental Setting and Preparation 3.2 Handover Tasks with Big Boxes 4 Robot Control and Data Aquisition 4.1 Pathplanning and Overview Software Components 4.2 Measurement of Kinematics 4.3 Preprocessing and Visualization of Kinematic Data 4.4 Human Upper Body Kinematic Model 4.5 Event Detection 4.6 Parameters to Describe the Human Pose During Handover 5 Results 6 Discussion 7 Conclusion and Future Work References Biomimetic Robots A Simulation Study for Evaluating the Role of Pre-tensioned Springs in 3 Pneumatic Artificial Muscle Driven Joint Mechanisms with Sliding Mode Controllers 1 Introduction 2 Dynamics of Robot Actuated by PAM 2.1 Dynamics of PAM 2.2 Dynamics of Three Types of Antagonistic Models 3 Sliding Mode Control for the Three Types of Structures 4 Validation and Results Analysis 4.1 Experiment Setting 4.2 Experiment Results 5 Discussion 6 Conclusion and Future Work References Effect of Tilted Ground on Muscle Activity in Human Sit-to-Stand Motion: Preliminary Result 1 Introduction 2 Method 2.1 Approach 2.2 Experimental Procedure 3 Result 4 Discussion 5 Conclusion References In-Hand Object Recognition for Sensorized Soft Hand 1 Introduction 2 Related Works 2.1 Sensorized Soft Actuators 3 Object Recognition 4 Methods 4.1 System Overview 4.2 Neural Network 4.3 Data Collection 5 Results and Discussion 6 Conclusion References Scaffolded Learning of In-place Trotting Gait for a Quadruped Robot with Bayesian Optimization 1 Introduction 2 Methods 2.1 Quadruped Gait Controller 2.2 Bayesian Optimization 2.3 Experimental Setup 3 Results and Discussion 4 Conclusion References GradNet: A Viscosity Gradient Approach to Achieve Dexterity in Soft Pneumatic Actuators 1 Introduction 2 Related Work 3 Methodology 3.1 Mixing of Elastomers 3.2 Designing of Non-Parallel Surface Actuator with Fiber Reinforcements 3.3 Fabrication of Gradiented and Non-gradriented Actuators 4 Comparison Between SGA and SGGA 4.1 Quantitative Testing Using Highlighted Markers 5 Conclusion and Future Work References How to Tune Humanoid Walking Parameters for Better Performance 1 Introduction 2 Problem Position 2.1 Rigid Link Leg 2.2 Spring-based Leg 3 Model Outline 3.1 Inversion 3.2 Critical Points 4 The Gait Parameters 4.1 Leg Design 4.2 Building a Leg 5 Spring-based Legs 5.1 First Setup 5.2 Second Setup 6 Conclusions References Machine Learning TridentNet: A Conditional Generative Model for Dynamic Trajectory Generation 1 Introduction 2 Related Work 2.1 Autonomous Navigation Frameworks 2.2 Scene Representations 2.3 Trajectory Forecasting 3 TridentNet CGM 3.1 Approach 3.2 Global Plan 3.3 Local Scene Representation 3.4 Conditional Variational Autoencoder 3.5 Dynamic Path Generation Using TridentNet 4 Data and Experiments 4.1 Calibration 4.2 Dataset and Ground Truth Labels 4.3 Experiments 5 Conclusion and Future Work References Adaptive Eligibility Traces for Online Deep Reinforcement Learning 1 Introduction 2 Preliminaries 2.1 Reinforcement Learning 2.2 Eligibility Traces 2.3 Deep Neural Networks 3 Adaptive Eligibility Traces 3.1 Problem and Solution 3.2 Policy Density Ratio in Online Learning 3.3 Decaying Factor According to Output Divergences 4 Simulations 4.1 Network Architecture 4.2 Benchmark Tasks 4.3 Results 5 Conclusion References Ensemble of LSTMs and Feature Selection for Human Action Prediction 1 Introduction 2 Proposed Human Action Prediction 2.1 Ensembles of LSTMs for Action Prediction 2.2 Feature Selection 3 Experimental Results 3.1 Dataset and Training 3.2 Experimental Results 4 Conclusion and Future Work References Control Analysis of Exploration in Swarm Robotic Systems 1 Introduction 2 Related Work 3 Problem Statement 4 Proposed State Representation 5 Derivation of Hitting Time Distributions 5.1 Hitting Time Distribution: Single Robot 5.2 Hitting Time Distribution: Multiple Robots 6 Experimental Setup 7 Results 8 Conclusions and Future Work References Reinforcement Learning at the Cognitive Level in a Belief, Desire, Intention UAS Agent 1 Introduction And Background 1.1 The Role of Reinforcement Learning 1.2 Complexity and Cognitive Structure 1.3 Complexity and Airspace Structure 2 Learning and BDI 3 Experiments 3.1 Environment Model 3.2 Actions 3.3 Transition Probabilities and Rewards 3.4 Policy Selection 4 DISCUSSION 5 Conclusion and Future Research References Action Feasibility Learning with Cell-Based Multi-Object Representation for Task and Motion Planning 1 Introduction 2 Problem Definition 3 Proposed Approach 3.1 Anchor-Cell-Based Scene Representation 3.2 Network Architecture 3.3 Data Collection Process 4 Results 5 Conclusion References Improved Driving Stability of Segway Using Two Serial Elastic Actuators 1 Introduction 2 Segway Design 2.1 Segway Model Comparison 2.2 Use of Series Elastic Actuator 3 Segway Modeling 4 Precision Driving Algorithm 5 Experiment 6 Conclusion References Optimal Grasping Control System for Performing Precision Indirect Teaching of Robot Hand 1 Introduction 2 System Configuration 2.1 Indirect Teaching System Configuration 2.2 Robot Hand Control System Design 2.3 Configuration of Contact Sensing Module 3 Fuzzy Control System 4 Experiment 5 Conclusion References Introduction of Damping Control for Obstacle Avoidance in Direct-Contact Cobotics Operations 1 Introduction 2 Robot Arm Using Torque Control 2.1 Jaco2 by Kinova 2.2 Application Situation 3 Damping Control Implementation 3.1 Actuator Damping 3.2 Damping Modification to Warn the Operator of an Obstacle 3.3 Implementation of Damping Command 4 Experiments 4.1 Experimental Setup 4.2 Experiment Sequence 5 Results and Discussion 5.1 Performance of the System from the Registered Data 5.2 Results from the Questionnaire 6 Conclusion References Shape Control of Elastic Objects Based on Implicit Sensorimotor Models and Data-Driven Geometric Features 1 Introduction 2 Methodology 2.1 Overview 2.2 Online Identification of Shape Feature 2.3 Derivation of Analytical Jacobian Matrix 2.4 Designing Controller 3 Case of Study: Bending Based on Spatial Arc 3.1 Analysis of the Task 3.2 Analytical Jacobian Matrix 4 Experiments and Results 5 Conclusion References Robotic Arm Control and Task Training Through Deep Reinforcement Learning 1 Introduction 2 State of the Art 3 Theoretical Background 3.1 Problem Formulation 3.2 Theoretical Analysis of Selected Algorithms 4 Robot Modeling 5 Assigned Tasks 5.1 Random Target Reaching 5.2 Pick&Place 6 Training Setup 7 Algorithms Comparison and Results 7.1 Random Target Reaching 7.2 Pick&Place 7.3 Real-World Experiments 8 Conclusions and Future Work References A Marker Based Optical Measurement Procedure to Analyse Robot Arm Movements and Its Application to Improve Accuracy of Industrial Robots 1 Introduction 2 Robot Control and Data Acquisition 2.1 Experimental Setting and Preparation 2.2 Pathplanning 2.3 Measurement of Kinematics 2.4 Phase Detections 2.5 Overview About the Software Components 3 Experiments 3.1 Fitting of Coordinate Systems and Determination of the Robots Position Repeatability (Precision) 3.2 Determination of Denavit-Hartenberg Parameters 4 Results 5 Discussion 6 Conclusion and Future Work References A Planning Domain Definition Language Generator, Interpreter, and Knowledge Base for Efficient Automated Planning 1 Introduction 1.1 Contribution 2 Related Work 3 Theoretical Background: Classical Automated Planning and PDDL Formulations 4 Our Proposal 4.1 PDDL Generator and Interpreter 4.2 Knowledge Base and Semantic Concepts 5 Empirical Evaluation 5.1 Blocks World 5.2 Assembly 6 Conclusions and Future Work References Weighted Shared-Autonomy with Assistance-to-Target and Collision Avoidance for Intelligent Assistive Robotics 1 Introduction 2 Methods 2.1 Assistance-to-Target 2.2 Collision Avoidance 2.3 Weighted Shared-Autonomy 3 Experiment 4 Results 5 Conclusions References Automatic Grasp Pose Generation for Parallel Jaw Grippers 1 Introduction 2 Related Work 2.1 Basic Contact Principles 2.2 Grasp Synthesis 3 Automatic Grasp Pose Generation 3.1 Grasp Synthesis 3.2 Grasp Clustering 4 Evaluation 4.1 Qualitative Analysis 4.2 Experiments in a Real-World Robot Cell 4.3 Discussion 5 Conclusions and Future Work References Simulation and Control of Deformable Autonomous Airships in Turbulent Wind 1 Introduction 2 State of the Art 3 Methodology 3.1 Simulation 3.2 Control 3.3 Middleware 3.4 Robotic Hardware 3.5 HITL 4 Experiments and Results 4.1 Simulation Experiments 4.2 Real-World Experiment 4.3 Discussion 5 Limitations 6 Summary References Computer Vision Visualization of Dump Truck and Excavator in Bird's-eye View by Fisheye Cameras and 3D Range Sensor 1 Introduction 2 Proposed Method 2.1 Overview 2.2 Definition of the Position and Pose in Three Coordinate Systems 2.3 Estimation of the Position, Pose, and Joint Poses of the Excavator 2.4 Estimation of the Position and Pose of the Dump Truck 2.5 Display of 3D Models in Bird's-Eye View 2.6 System Architecture Diagram of the Proposed Method 3 Experiment 3.1 Experimental Setting 3.2 Experimental Result 4 Conclusion References Dynamic Object Removal from Unpaired Images for Agricultural Autonomous Robots 1 Introduction 2 Related Work 3 Method 3.1 Loss Function 3.2 Network Architecture 3.3 Training Details 4 Dataset 5 Evaluation 5.1 Quantitative Analysis 5.2 Qualitative Analysis 6 Conclusion References Future Image Prediction for Mobile Robot Navigation: Front-Facing Camera Versus Omni-Directional Camera 1 Introduction 2 Related Work 2.1 Image Prediction Using Deep Learning 2.2 Deep Visual Navigation 3 Learning Future Image Prediction 3.1 Network Structure 3.2 Training Procedure 3.3 Network Implementation and Training Details 4 Image Prediction Experiment 4.1 Simulation 4.2 Real-world Experiment 5 Navigation Experiment 6 Conclusions and Future Work References Improving Embedded Convolutional Object Detectors Through Low-res Training 1 Introduction 2 Methods 2.1 YOLOv3 Tiny 2.2 MobileNet SSD 2.3 Light Object Detection Framework 2.4 Experimental Setup 3 Results 4 Discussion 4.1 Speed Superiority from YOLO 4.2 SSD Problems with Low Resolutions 4.3 Optimal Resolution 4.4 Choice Considerations 5 Conclusion References Fusion of Radar- and Lidar-Data for Object-Tracking-Applications at Feature Level 1 Introduction 1.1 Background 1.2 Kalman-Filter 1.3 Fusion 2 Object-Tracking 3 Fusion 3.1 Concept of Fusion Strategy 3.2 Classification of Fusion Strategy 4 Multi-sensor-System 4.1 Arrangement of Sensors 4.2 Synchronization 5 Experimental Results 5.1 Conclusion 5.2 Outlook References Learning to Segment Human Body Parts with Synthetically Trained Deep Convolutional Networks 1 Introduction 2 Related Work 3 Body-Part Synthetic Data Generation 4 Shape-Aware Segmentation 5 Experiments 5.1 Experimental Setup 5.2 Performance Evaluation 5.3 Results 5.4 Ablation Study 6 Conclusions References Detection and Classification of Defects in Plastic Components Using a Deep Learning Approach 1 Introduction 2 Related Works 3 Materials and Methods 3.1 Data Acquisition 3.2 Data Processing 3.3 Data Segmentation 4 Results and Discussions 5 Conclusion and Future Works References Author Index