Robot Talk Episode 170: Daniela Shchegelskaya on Reliable Robots in Safety-Critical Work

In Robot Talk Episode 170, host Claire speaks with Daniela Shchegelskaya of the University of Twente about the challenge of making robots operate reliably in safety-critical environments. The episode examines what changes when autonomous systems move from controlled laboratories into settings where failures carry real operational consequences.

Shchegelskaya is an electronics and robotics engineer whose work spans autonomous systems, control engineering, and AI-driven robotics for industrial and safety-critical applications. Her engineering focus is on deploying real-world autonomous robots beyond laboratory conditions, with attention to reliability, connectivity, and system-level design.

Rather than treating reliability as a single component property, the discussion frames it as a system-wide concern. A robot that performs well in a demonstration may still struggle when networks are variable, diagnostics are handled remotely, and the surrounding infrastructure imposes strict operating constraints.

At Elimpus, Shchegelskaya works on integrating Boston Dynamics’ Spot for inspections in live power substations. That application is demanding because the robot operates where safety, uptime, and operational discipline are essential. The integration work includes autonomous navigation, network connectivity, remote diagnostics, data collection, and system-level reliability.

The episode places those tasks in a broader deployment context. Inspection robots in live substations must move through complex sites, maintain communications, support remote troubleshooting, and collect data that operators can trust. These elements are interdependent: connectivity problems can limit autonomy, and weak diagnostics can reduce confidence in the data a robot gathers.

The conversation also highlights the gap between laboratory performance and field reliability. In safety-critical environments, reliability is not only a question of hardware robustness. It also depends on system integration, network design, remote support, and clear operational procedures that define how the robot and human teams work together.

Shchegelskaya’s profile reflects a growing emphasis on deployment-focused robotics. The field needs engineers who can connect control algorithms and AI methods to the practical demands of industrial sites, where robots must work alongside existing infrastructure, maintenance schedules, and human operators.

For listeners, the discussion offers a grounded view of what it takes to move autonomous robots into real-world use. The emphasis is less on novelty and more on dependable operation: how a robot stays connected, how failures are diagnosed, how data is collected, and how the overall system earns trust in safety-critical work.

The episode also shows why connectivity and diagnostics deserve attention at the design stage. A robot that cannot be monitored or supported remotely is harder to deploy in a live substation, even if its navigation software performs well in testing. Remote diagnostics help teams understand faults, plan interventions, and maintain confidence in the system.

System-level reliability therefore becomes a practical engineering goal. It requires testing across hardware, software, networks, and operating procedures, plus a clear understanding of how failures propagate. For industrial robotics, that discipline is central to turning a capable robot platform into a dependable inspection tool.

Autonomous navigation is only one part of the challenge. In a live substation, the robot must follow routes safely, avoid restricted areas, and handle changes in the environment. Network connectivity adds another layer, because remote supervision and data transfer depend on stable communication links that may be constrained by industrial infrastructure.

Data collection is equally important. Inspection missions generate images, readings, and status information that must be organized and delivered to the right people. If the data is incomplete or hard to interpret, the robot’s value is limited. Reliability therefore includes the quality and usability of the information the system produces.

Shchegelskaya’s work at Elimpus illustrates how deployment teams combine robotics with operational engineering. The goal is not simply to demonstrate a robot walking through a facility. It is to create a repeatable inspection capability that fits the routines and safety requirements of the site.

The Robot Talk episode uses this example to explore the wider question of trust in autonomous systems. Trust grows when robots behave predictably, when problems are visible, and when operators have the information they need to respond. Those conditions are technical and organizational at the same time.