Designing an efficient haul truck dispatching in open-pit mining produces unique challenges when considering the stochastic, dynamic nature of a mine operation. Travel times, loading times, queue lengths, and equipment availability can fluctuate, particularly under changing environmental or operational conditions. While traditional methods can approximate this variation, updating these solutions is time-intensive and reactionary. Additionally, the greater effects of truck maintenance is often overlooked in favour of short-term production targets, which can reduce the fleet’s overall reliability. To achieve a truly intelligent dispatcher, both production and maintenance should be considered in tandem in real time.<\/p>\n\n\n\n
We recognize the interplay between these two problems, and in this paper present a more robust method to handle both objectives using deep reinforcement learning (DRL). We setup and allow this new dispatching model to interact with a simulated copper mine environment, assigning trucks to destinations based on real-time measures such as estimated grade, shovel queues, and equipment health. We generate “rewards” for the dispatching decisions based on efficiency; by maximizing the rewards received, the model learns to make more effective choices. By implementing several DRL variations that each enhance stochastic adaptation in different ways, we provide a suite of models with unique approaches to the dispatching problem.<\/p>\n\n\n\n
A key strength of our methodology lies in the real-time design of the mine environment. With access to parameters for both individual trucks and mine-wide operational status, the model can make decisions that account for immediate conditions while predicting upcoming changes (such as a backlogged shovel). The introduction of maintenance as an operational parameter capitalizes on the growing use of internal truck sensors to approximate an overall vehicle health score, allowing proactive upkeep to be considered with some confidence. The reward function is designed to reflect practical goals: maximize production, reduce waiting times, and perform regular maintenance when necessary. As random equipment failures can occur even with regular maintenance, the system learns to robustly handle a dynamic fleet size. Combined, these efforts produce a complex yet coherent system where DRL methods can excel.<\/p>\n\n\n\n
Our results indicate that the models outperform conventional heuristic approaches, such as shortest queue (SQ) and fixed scheduling (FS). Performance gains are observed in terms of increased production rates, reduced idle times, and fewer truck breakdown events, culminating in increased overall shift profits. Notably, the advantage of the DRL approach becomes more pronounced as the level of uncertainty increases, highlighting its suitability for environments where variability is a defining characteristic. An additional benefit of our models is the capacity for real-time adaptation. As training occurs with the risk of frequent equipment failures, the fully operational models can handle a dynamic fleet size with confidence.<\/p>\n\n\n\n
Our work here proves that DRL can provide an adaptive real-time solution to the dispatch problem, with a balance between immediate production returns and long-term fleet reliability. With further realism implemented in the mine simulation in future work, we take a step towards full deployment, to improve both efficiency and safety in open-pit mine operations. Find the full article in Engineering Optimization here: https:\/\/www.tandfonline.com\/doi\/abs\/10.1080\/0305215X.2025.2600329<\/a>. <\/p>\n","protected":false},"excerpt":{"rendered":" Designing an efficient haul truck dispatching in open-pit mining produces unique challenges when considering the stochastic, dynamic nature of a mine operation. Travel times, loading times, queue lengths, and equipment availability can fluctuate, particularly under changing environmental or operational conditions. While traditional methods can approximate this variation, updating these solutions is time-intensive and reactionary. Additionally,… <\/p>\n","protected":false},"author":8,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"neve_meta_sidebar":"default","neve_meta_container":"","neve_meta_enable_content_width":"","neve_meta_content_width":70,"neve_meta_title_alignment":"left","neve_meta_author_avatar":"","neve_post_elements_order":"[\"title\",\"meta\",\"content\",\"tags\",\"comments\"]","neve_meta_disable_header":"","neve_meta_disable_footer":"","neve_meta_disable_title":"","neve_meta_reading_time":"","footnotes":""},"categories":[1],"tags":[9,7,8,6],"class_list":["post-556","post","type-post","status-publish","format-standard","hentry","category-uncategorized","tag-deep-q-learning","tag-open-pit-dispatch","tag-optimization","tag-reinforcement-learning"],"_links":{"self":[{"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/posts\/556","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=556"}],"version-history":[{"count":3,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/posts\/556\/revisions"}],"predecessor-version":[{"id":559,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=\/wp\/v2\/posts\/556\/revisions\/559"}],"wp:attachment":[{"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=556"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=556"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ims.smithengineering.queensu.ca\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=556"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}