Autonomy: the Journey from Experience to Prescription

Read time: ~3min

Autonomy at scale promises to unlock real increases in efficiency as we learn from and adapt operations and maintenance to the ever-changing geology and physical environment at mine sites. The adoption of autonomous equipment is only one part of the equation. With ever-increasing economic, environmental and social pressures, autonomy in mining has quickly moved from a nice-to-have to a necessity for scaling and transformation. When faced with needing to do more with less impact, autonomy will help scale productivity, reduce human exposure to hazardous environments, optimize resource extraction and minimize environmental impact. Achieving meaningful autonomy at scale will require system level changes to how we work.

The journey to autonomy can be thought of as the change from experience and rules to predictions and prescriptions.

Experience

Our industry is founded upon experience. Experience has taught us what to do, and what not to do, in order to keep people safe, solve problems as they arise, and design and build operations that can cope with high variability. This experience is different from person to person, is limited to what the person has been exposed to, and is heavily influenced by risk mitigation, and based on our very limited sensory perception. Experience tells us not to touch exposed wires.

Rules

The codification of experience is rules. Rules are coded experiences. Engineers calculate, superintendents perform, investigations occur, learning is limited to what can be applied next time in a similar situation, and edge cases are generally lost, or if serious enough, encoded into an additional rule. The rise of rules has enabled us to create organisations and scale. Rules do not learn from other rules, nor do they “know” what a situation is. Rules are boundaries to force action in a known situation and limit action to prevent failure. Rules provide known, predetermined limits on personal experience when making decisions Expert systems such as control systems and sources of truth drive business processes, limit risk and enable decision making.

Analytics

The examination of data in arrears to improve rules or make specific decisions is analytics. Analytics extends rules beyond human cognition and beyond individual experience. The rise of analytics in recent years is directly due to the decrease in cost of data acquisition, storage and compute (IOT & cloud in particular). Analytics introduces structured learning as a capability and has begun to shape rules and provide inputs to decisions. Machine Learning (ML) optimisations are often now used as a form of analytics that can solve problems that are extremely difficult to solve by brute force/inspection. Decisions about the future are still generally made by humans, and analytics outputs are generally compared to experience and existing rules (the past).

Predictions

Predictions and optimisations are the first step from in arrears decision making to future looking decisions. The distinction is subtle but critical - predictions are probabilistic, and are not necessarily based on linear extrapolations of historic data. Predictions are just that - predictions. That are not calculations. Calculations are forms of rules, that come from analytics, and probabilistic calculations are forms of analytics, not predictions. Optimisations that consider predictions are a special class of hybrid analytics that includes future prediction with historic data to deliver improved (more usable) optimisations than analytics alone. The use, or adoption, of predictions is very use case specific; some predictions are much lower risk than others, and some forms of action would be highly risky if a low probability prediction was used without regard to analytics, rules or experience.

Prescriptions

The codification of predictions, or translation of predictions into rules that are not wholly based on experience, is called prescriptions. Prescriptions enable systems to take action based on changes in current data, which in turn create a trigger for action, because the current data has changed the prediction of what will happen next. The system “learns” and takes action, or recommends action to a human, based on a prediction of the future.

Some predictions can be folded into expert systems (rules) and improve the efficiency of those rules based systems, by selecting which rule to apply, or by changing existing rules within defined parameters, or by creating new rules based on the predictions and use case. We can think of these as “adaptive rules” or “dynamic rules”, and the rules based system can have static rules which in turn examine the predicted dynamic rules, to provide the limitations that we require on any dynamic decision making process (e.g. rules limit and guide our experience).

From Rules to Prescriptions

The parallels between expert/static rules based on experience, and learning/adaptive rules based on data, are not coincidental. Human cognition is finite, and already vastly overwhelmed by the availability of data and influence of opinions and emotions (personal and others experience). The journey from static rules that align with our experience to adaptive rules based on predictions that challenge our experience is the journey that all organisations must make if they are to embrace the autonomy opportunity. This change is not an IT, data science, analytics, or ML challenge. It is a system level challenge on how we decide, manage, incentivise, support, train, report, and decide what to do next, at the individual, team, site and organizational levels. The rules to prescription journey are the no.1 challenge and opportunity for the 21st mining company, and every other industry that operates on domain specific experience and expert rules.