[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84106-en":3,"doc-seo-84106-105":28,"detail-sidebar-cat-0-en-105":89},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":13,"seo_description":14,"update_tm":26,"read_time":27},84106,1099514067438,"River Wang","https://ap-avatar.wpscdn.com/avatar/100002539ee87300030?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780474512215547542",8,"Research & Report","Learning to Throw Objects Safely in Multi-Obstacle Environments","Robotic throwing supports rapid object placement beyond a robot’s immediate workspace, yet dependable throwing in cluttered scenes is still insufficiently addressed. This work studies throwing an object into a target basket while avoiding randomly placed obstacles. A potential field state representation compactly encodes basket attraction and obstacle repulsion on a fixed grid, allowing reinforcement learning policies to generalize across varying obstacle counts and layouts. Policies are initialized via kinesthetic demonstrations and trained in simulation using SAC, DDPG, and TD3; SAC is most consistent. Real-robot tests with unseen objects achieve up to 90% success and robust sim-to-real transfer.","Learning to Throw Objects Safely in Multi-Obstacle Environments  \nMohammadreza Kasaei2 , Klemen Voncina 1 , and Hamidreza Kasaei 1  \narXiv :2607 .06388v 1 [ cs .RO] 7 Jul 2026  \nAbstract—Robotic throwing enables fast and efficient object placement beyond the robot’s immediate workspace, but reliable throwing in cluttered environments remains underexplored. Existing approaches, such as TossingBot, learn throwing strategies from visual input but assume obstacle-free settings. In this paper, we address the problem of throwing objects into a target basket while avoiding obstacles placed randomly in the scene. We introduce a potential field state representation that compactly encodes both basket attraction and obstacle repulsion on a fixed-size grid, enabling reinforcement learning (RL) policies to generalize across arbitrary numbers and configurations of obstacles. The policy is initialized from kinesthetic demonstrations and optimized in simulation using three state-of-the-art RL algorithms (SAC, DDPG, TD3). Among these, SAC achieves the most consistent performance across scenarios. We compare the potential field representation against explicit state encodings and demonstrate that it achieves higher success rates and better scalability to unseen obstacle configurations. Real-robot experiments with unseen throwable objects confirm robust sim-to-real transfer, achieving up to 90% success in cluttered scenes. These results demonstrate that PFR provides a practical and robust representation for safe and efficient robotic throwing in unstructured environments. A video showcasing our experiments is available at: [https://youtu.be/ZZnJf8ua2dE](https://youtu.be/ZZnJf8ua2dE)  \nI. INTRODUCTION In automated distribution and sorting logistics systems, speed and efficiency play a critical role [1] . Developing robots with the ability to swiftly pick up and throw objects can significantly enhance throughput and operational efficiency. Throwing, in particular, enables fast and precise object placement, reduces the need for complex robotic traversal, and extends the robot’s effective workspace [2] .  \nBy allowing robots to toss objects instead of carrying them, throughput can be improved in warehouses, automated fulfillment centers, and waste-sorting systems.  \nRecent works have demonstrated the feasibility of robotic throwing. For example, TossingBot [3] learns grasping and throwing policies from visual input, but assumes obstaclefree environments. Other approaches [4], [5] rely on handcrafted or analytic motion kernels, limiting adaptability in cluttered settings. However, practical environments rarely allow direct throwing trajectories: obstacles such as walls, bins, or other items must be considered to ensure successful placement.  \nAn illustrative example is shown in Fig. 1. Suppose the robot is asked to place an object into the gray basket.  \n1 Klemen Voncina and Hamidreza Kasaei are with the Department of Artificial Intelligence, University of Groningen, The Netherlands. Emails: [k.voncina@student.rug.nl](k.voncina@student.rug.nl), [hamidreza.kasaei@rug.nl](hamidreza.kasaei@rug.nl)  \n2 Mohammadreza Kasaei is with the School of Informatics, University  \nof Edinburgh, [UK. Email: m.kasaei@ed.ac.uk](UK. Email: m.kasaei@ed.ac.uk)  \nFig. 1: Overview of our throwing task. The robot must decide between a slow object-handover or a faster throwing action. Since obstacles may block the trajectory, we treat the problem as a Safe Reinforcement Learning challenge, where the goal is to achieve reliable throws while avoiding collisions.  \nSince the basket lies outside the maximum kinematic reach of the left arm, a simple placement action is not feasible. One option would be to perform a hand-over to the right arm and then place the object into the basket, but this is a relatively slow process and reduces system throughput. Alternatively, the robot can throw the object directly into the basket. However, this introduces additional challenges: the robot mu","cbCaiqqtIqf3XFMC","https://ap.wps.com/l/cbCaiqqtIqf3XFMC","pdf",5575243,1,"English","en",105,"# Introduction\n## Robotic throwing in logistics\n## Safe reinforcement learning formulation\n## Potential field representation\n## Experimental setup and contributions","[{\"question\":\"What is the main problem addressed in this paper?\",\"answer\":\"The paper addresses reliable robotic throwing into a target basket while avoiding multiple obstacles randomly placed in the scene.\"},{\"question\":\"How does the proposed potential field state representation help reinforcement learning generalize?\",\"answer\":\"It encodes basket attraction and obstacle repulsion on a fixed-size grid, producing a fixed-dimensional input that scales without depending on the number of obstacles.\"},{\"question\":\"Why are kinesthetic demonstrations used, and which RL algorithm performs best?\",\"answer\":\"Kinesthetic teaching initializes the throwing motion with a safe kernel to reduce unsafe exploration during early RL training; among SAC, DDPG, and TD3, SAC delivers the most consistent performance across scenarios.\"}]",1784192872,20,{"code":4,"msg":29,"data":30},"ok",{"site_id":23,"language":22,"slug":31,"title":13,"keywords":32,"description":14,"schema_data":33,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":26},"learning-to-throw-objects-safely-in-multi-obstacle-environments","",{"@graph":34,"@context":83},[35,52,66],{"@type":36,"itemListElement":37},"BreadcrumbList",[38,42,46,49],{"item":39,"name":40,"@type":41,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":43,"name":44,"@type":41,"position":45},"https://docshare.wps.com/document/","Document",2,{"item":47,"name":12,"@type":41,"position":48},"https://docshare.wps.com/document/research-report/",3,{"item":50,"name":13,"@type":41,"position":51},"https://docshare.wps.com/document/learning-to-throw-objects-safely-in-multi-obstacle-environments/84106/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":22,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":39,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"What is the main problem addressed in this paper?","Question",{"text":73,"@type":74},"The paper addresses reliable robotic throwing into a target basket while avoiding multiple obstacles randomly placed in the scene.","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"How does the proposed potential field state representation help reinforcement learning generalize?",{"text":78,"@type":74},"It encodes basket attraction and obstacle repulsion on a fixed-size grid, producing a fixed-dimensional input that scales without depending on the number of obstacles.",{"name":80,"@type":71,"acceptedAnswer":81},"Why are kinesthetic demonstrations used, and which RL algorithm performs best?",{"text":82,"@type":74},"Kinesthetic teaching initializes the throwing motion with a safe kernel to reduce unsafe exploration during early RL training; among SAC, DDPG, and TD3, SAC delivers the most consistent performance across scenarios.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,118,121,125,128,132],{"id":20,"doc_module":4,"doc_module_name":44,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":45,"doc_module":4,"doc_module_name":44,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":44,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":44,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":44,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":44,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":44,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":44,"category_name":123,"show_sort_weight":27,"slug":124},9,"Religion & Spirituality","religion-spirituality",{"id":27,"doc_module":4,"doc_module_name":44,"category_name":126,"show_sort_weight":27,"slug":127},"World Cup","world-cup",{"id":129,"doc_module":4,"doc_module_name":44,"category_name":130,"show_sort_weight":129,"slug":131},10,"Lifestyle","lifestyle",{"id":133,"doc_module":4,"doc_module_name":44,"category_name":134,"show_sort_weight":104,"slug":135},19,"General","general"]