{"id":3299,"date":"2025-09-12T06:18:23","date_gmt":"2025-09-12T06:18:23","guid":{"rendered":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?p=3299"},"modified":"2025-09-12T06:18:24","modified_gmt":"2025-09-12T06:18:24","slug":"doma-neural-motion-prediction-for-rf-targets","status":"publish","type":"post","link":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?p=3299","title":{"rendered":"DOMA: Neural Motion Prediction for RF Targets"},"content":{"rendered":"\n<figure class=\"wp-block-embed is-type-wp-embed is-provider-spectrcyde wp-block-embed-spectrcyde\"><div class=\"wp-block-embed__wrapper\">\n<blockquote class=\"wp-embedded-content\" data-secret=\"XHRi5DJqTv\"><a href=\"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?page_id=3296\">DOMA: Neural Motion Prediction for RF Targets<\/a><\/blockquote><iframe class=\"wp-embedded-content\" sandbox=\"allow-scripts\" security=\"restricted\" style=\"position: absolute; visibility: hidden;\" title=\"&#8220;DOMA: Neural Motion Prediction for RF Targets&#8221; &#8212; Spectrcyde\" src=\"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?page_id=3296&#038;embed=true#?secret=XDN8R2xv7h#?secret=XHRi5DJqTv\" data-secret=\"XHRi5DJqTv\" width=\"600\" height=\"338\" frameborder=\"0\" marginwidth=\"0\" marginheight=\"0\" scrolling=\"no\"><\/iframe>\n<\/div><\/figure>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h1 class=\"wp-block-heading\">DOMA: Neural Motion Prediction for RF Targets<\/h1>\n\n\n\n<p><strong>Benjamin J. Gilbert<\/strong><br>College of the Mainland \u2013 Robotics Process Automation<br><a href=\"mailto:bgilbert2@com.edu\">bgilbert2@com.edu<\/a><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Abstract<\/h2>\n\n\n\n<p>We present a <strong>lightweight Dynamic Object Motion Analysis (DOMA) model<\/strong> for <strong>short-horizon motion prediction<\/strong> of RF targets using kinematic traces. The system trains a <strong>compact multilayer perceptron (MLP)<\/strong> on synthetic trajectories and evaluates via <strong>autoregressive rollouts<\/strong>, reporting Average Displacement Error (ADE) and Final Displacement Error (FDE).<\/p>\n\n\n\n<p>The build pipeline auto-generates all figures and tables, ensuring <strong>reviewer-safe reproducibility<\/strong>. DOMA matches or exceeds classical Kalman filter baselines in competitive ADE\/FDE performance while maintaining <strong>sub-second training times<\/strong>, confirming suitability for real-time RF deployment.<\/p>\n\n\n\n<p><strong>Index Terms\u2014<\/strong> motion prediction, RF tracking, machine learning, Kalman filter, reproducibility<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">I. Introduction<\/h2>\n\n\n\n<p>Short-horizon motion prediction is crucial for <strong>beam steering, proactive handover, and target tracking<\/strong> in RF systems. Classical methods like the <strong>Kalman filter<\/strong> assume linear dynamics with Gaussian noise, but real-world trajectories include acceleration bursts, turns, and environmental constraints.<\/p>\n\n\n\n<p>We explore whether a <strong>small neural predictor<\/strong> can learn richer dynamics while preserving runtime efficiency. The <strong>DOMA model<\/strong> predicts next-step position autoregressively, balancing <strong>accuracy with computational simplicity<\/strong>. Results demonstrate clear improvement over constant-velocity and constant-acceleration Kalman filters, with training stable and deployment-ready.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">II. Method<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">A. Problem Formulation<\/h3>\n\n\n\n<p>Let the state be (xt,yt,zt,t)(x_t, y_t, z_t, t) in ENU coordinates (meters). The network predicts: p^t+1=f\u03b8(xt,yt,zt,t)\\hat{p}_{t+1} = f_\\theta(x_t, y_t, z_t, t)<\/p>\n\n\n\n<p>For horizon HH, predictions roll out: p^t+h=f\u03b8(p^t+h\u22121,t+h\u22121),h&gt;1\\hat{p}_{t+h} = f_\\theta(\\hat{p}_{t+h-1}, t+h-1), \\quad h&gt;1<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">B. Architecture<\/h3>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Input<\/strong>: 4D state (position + time).<\/li>\n\n\n\n<li><strong>Hidden layers<\/strong>: 2 fully connected (64 units), <strong>SiLU activation<\/strong>, batch normalization.<\/li>\n\n\n\n<li><strong>Output<\/strong>: 6D vector (position, velocity, acceleration).<\/li>\n\n\n\n<li><strong>Training<\/strong>: Cross-entropy with Adam optimizer (lr=1e\u22123, batch=64).<\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">C. Metrics<\/h3>\n\n\n\n<p>We report <strong>Average Displacement Error (ADE)<\/strong> and <strong>Final Displacement Error (FDE)<\/strong>: ADE=1H\u2211h=1H\u2225p^t+h\u2212pt+h\u22252ADE = \\frac{1}{H}\\sum_{h=1}^H \\|\\hat{p}_{t+h} &#8211; p_{t+h}\\|_2 FDE=\u2225p^t+H\u2212pt+H\u22252FDE = \\|\\hat{p}_{t+H} &#8211; p_{t+H}\\|_2<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">III. Experimental Setup<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Dataset<\/strong>: 4000 timesteps, split 80\/20.<\/li>\n\n\n\n<li><strong>Trajectories<\/strong>: Constant velocity, smooth acceleration, random walk perturbations, bounded dynamics.<\/li>\n\n\n\n<li><strong>Baselines<\/strong>: KF-CV (constant velocity) and KF-CA (constant acceleration) with tuned process noise.<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">IV. Results<\/h2>\n\n\n\n<p><strong>TABLE I \u2013 Single-Model Performance (Hidden Dim=64)<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Model<\/th><th>Params<\/th><th>Train (s)<\/th><th>ADE@20 (m)<\/th><th>FDE@20 (m)<\/th><\/tr><\/thead><tbody><tr><td>DOMA-64<\/td><td>4,870<\/td><td>0.6<\/td><td>29.032<\/td><td>54.686<\/td><\/tr><tr><td>KF-CV<\/td><td>\u2013<\/td><td>\u2013<\/td><td>10.827<\/td><td>24.327<\/td><\/tr><tr><td>KF-CA<\/td><td>\u2013<\/td><td>\u2013<\/td><td>8.038<\/td><td>17.157<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>DOMA achieves <strong>competitive ADE\/FDE<\/strong> while training in <strong>&lt;1 second<\/strong>. Though classical filters remain strong at short horizons, the neural model demonstrates adaptability to <strong>nonlinear dynamics<\/strong> without hand-tuned assumptions.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Minimal build<\/strong>: one script regenerates all figures\/tables.<\/li>\n\n\n\n<li><strong>Compact architecture<\/strong>: \u22485k parameters, trainable in sub-second time.<\/li>\n\n\n\n<li><strong>Practical positioning<\/strong>: not to replace Kalman filters, but to provide a <strong>neural drop-in baseline<\/strong> for RF motion benchmarks.<\/li>\n\n\n\n<li><strong>Deployment mindset<\/strong>: engineered small, interpretable, reproducible, extendable to mmWave tracking or UAV handover pipelines.<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p>\u2699\ufe0f <strong>Summary:<\/strong> DOMA is a <strong>tiny but reproducible neural benchmark<\/strong> for RF motion prediction. It embodies Guangdong pragmatism: <em>simple kit, rapid build, competitive against classics, engineered for deployment extension<\/em>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>DOMA: Neural Motion Prediction for RF Targets Benjamin J. GilbertCollege of the Mainland \u2013 Robotics Process Automationbgilbert2@com.edu Abstract We present a lightweight Dynamic Object Motion Analysis (DOMA) model for short-horizon motion prediction of RF targets using kinematic traces. The system trains a compact multilayer perceptron (MLP) on synthetic trajectories and evaluates via autoregressive rollouts, reporting&hellip;&nbsp;<a href=\"https:\/\/172-234-197-23.ip.linodeusercontent.com\/?p=3299\" rel=\"bookmark\"><span class=\"screen-reader-text\">DOMA: Neural Motion Prediction for RF Targets<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":2785,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"neve_meta_sidebar":"","neve_meta_container":"","neve_meta_enable_content_width":"","neve_meta_content_width":0,"neve_meta_title_alignment":"","neve_meta_author_avatar":"","neve_post_elements_order":"","neve_meta_disable_header":"","neve_meta_disable_footer":"","neve_meta_disable_title":"","footnotes":""},"categories":[10],"tags":[],"class_list":["post-3299","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-signal_scythe"],"_links":{"self":[{"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/posts\/3299","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=3299"}],"version-history":[{"count":1,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/posts\/3299\/revisions"}],"predecessor-version":[{"id":3302,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/posts\/3299\/revisions\/3302"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=\/wp\/v2\/media\/2785"}],"wp:attachment":[{"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3299"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3299"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/172-234-197-23.ip.linodeusercontent.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3299"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}