Key Takeaways
- Two hospitality robots, Luna and Pepper, at Yotel Miami stopped functioning after a year of service, leaving the hotel without a viable technical solution.
- A multidisciplinary team of five Miami Dade College students revived the robots using a combination of hardware expertise, software skills, language training, mapping knowledge, and hospitality insight.
- Artificial intelligence played a pivotal role in diagnosing a laser‑related power issue and in creating a bilingual, Miami‑accented language model for the robots.
- The project not only restored the robots’ original concierge functions but also enhanced them, demonstrating how academic‑industry collaboration can solve emerging‑technology problems.
- The experience equipped the students with transferable skills applicable to logistics, hospital delivery, warehouse automation, and other sectors that rely on similar robotic platforms.
Background of the Yotel Robots
Yotel’s downtown Miami location deployed two service robots named Luna and Pepper to act as digital concierges for guests. According to General Manager Joseph LaFleur, the robots were programmed to deliver items such as toiletries, bath products, and dairy supplies directly to guest rooms, while also offering localized recommendations for dining, entertainment, and transportation. Their presence was intended to streamline front‑desk operations and provide a novel, tech‑savvy experience for visitors. For roughly a year, the robots performed these tasks reliably, becoming a recognizable part of the hotel’s brand identity.
The Problem Emerges
After approximately twelve months of continuous operation, Luna and Pepper began to malfunction. The exact nature of the failure was not immediately apparent; the robots would not power on, and their diagnostic lights remained inert. LaFleur recalled that the hotel’s technical team attempted standard troubleshooting procedures—checking connections, resetting firmware, and replacing obvious wear parts—but none of these actions restored functionality. The issue appeared to be rooted in a more obscure subsystem, prompting LaFleur to describe the situation as hitting “roadblocks” that no local tech provider could overcome.
Search for Solutions
Faced with an impasse, Yotel reached out to various robotics vendors, service companies, and even university research labs across South Florida. Despite the growing popularity of service robots in hospitality, none of the contacted entities possessed the specific expertise or documentation required to diagnose the Luna/Pepper platform. The hotel’s General Manager expressed frustration, noting that the technology was still relatively niche, and support networks were underdeveloped. This deadlock prompted Yotel to look beyond commercial channels and consider academic partnerships as a potential source of innovative problem‑solving.
Student Team Assembly
Miami Dade College’s West Campus responded to Yotel’s plea by assembling a group of five students under the guidance of instructor Philip Colodetti. The team was deliberately interdisciplinary: Patricia Hurtado brought a background in hospitality management; Noel Rodriguez and Samuel Suarez contributed hardware and software engineering expertise; Daniel Pena Calleja specialized in linguistics and natural‑language processing; and Liennis Montiel focused on geographic information systems and spatial mapping. This diverse skill set was intended to mirror the multifaceted challenges posed by reviving a service robot that needed mechanical, linguistic, and navigational competence.
Leveraging Individual Strengths
Colodetti emphasized that the project’s success hinged on allowing each student to apply their strongest abilities while learning from one another. Hurtado’s familiarity with hotel guest expectations helped the team prioritize features that would enhance the concierge experience, such as polite greetings and timely delivery notifications. Rodriguez and Suarez tackled the electrical and firmware layers, using oscilloscopes and logic analyzers to trace signal pathways. Pena Calleja examined the robots’ speech synthesis modules, identifying language packs and pronunciation quirks. Montiel undertook the task of creating accurate floor‑plan maps of the Yotel property, which would later enable the robots to navigate autonomously between the lobby, elevators, and guest rooms.
Initial Diagnosis and AI Assistance
The first technical hurdle was simply getting the robots to power on. Rodriguez recounted hours of manual testing—checking battery voltages, inspecting motor drivers, and verifying sensor feedback—without success. Frustrated by the lack of clear documentation, he turned to an artificial‑intelligence‑based diagnostic tool that could analyze patterns in error logs and suggest probable fault locations. The AI highlighted a malfunction in the robot’s lidar‑based laser system, which is used for obstacle detection and alignment during startup. Armed with this insight, Rodriguez replaced a faulty laser diode and recalibrated the associated circuitry, finally observing the robots’ boot sequences initiate.
Mapping the Robot’s System
With power restored, the team faced the challenge of integrating the robots’ navigation stack with the Yotel environment. Montiel explained that the robots relied on a proprietary simultaneous localization and mapping (SLAM) framework that differed from open‑source alternatives commonly taught in robotics courses. Because none of the students had prior exposure to this specific SLAM implementation, they adopted an iterative trial‑and‑error approach: they logged sensor data during manual walks through the hotel, fed the information into the mapping algorithm, and refined the map until the robots could reliably localize themselves within the lobby and corridor network. After several weeks of refinement, the students succeeded in generating a complete, occupancy‑grid map of the entire Yotel property.
Developing the AI Language Model
Parallel to the navigation work, Pena Calleja and Hurtado focused on endowing Luna with a natural, bilingual voice capable of engaging guests in both English and Spanish. Using a transformer‑based language model fine‑tuned on hospitality‑specific dialogues, they trained the system to recognize common guest requests—such as asking for extra towels, recommending nearby cafés, or providing directions to the beach. To impart a distinctive “Miami vibe,” the team incorporated regional speech patterns, colloquialisms, and a warm, upbeat prosody into the model’s output. The resulting voice enabled Luna to switch fluidly between languages while maintaining a congenial tone that matched the hotel’s brand personality.
Outcome and Impact
In just under two months, the student team had not only resurrected Luna and Pepper but also upgraded their capabilities beyond the original specifications. The robots now reliably delivered guest amenities, offered localized recommendations in two languages, and navigated the hotel’s layout with smooth, obstacle‑avoiding motion. LaFleur noted that guest feedback had become overwhelmingly positive, with many visitors commenting on the novelty and usefulness of the robotic concierge. The project also demonstrated a cost‑effective alternative to expensive vendor service contracts, showing that internal academic talent could address high‑tech problems swiftly and affordably.
Broader Implications for Industry
Colodetti highlighted that the underlying robotic platform used by Luna and Pepper is identical to systems deployed in logistics centers, hospital delivery fleets, and Amazon‑style warehouses. By mastering the hardware interfaces, SLAM navigation, and natural‑language modules specific to this platform, the students acquired a skill set directly transferable to multiple high‑growth sectors. He argued that such cross‑disciplinary projects prepare graduates for a future where robots will increasingly perform routine tasks across healthcare, retail, and manufacturing, and where the ability to adapt, troubleshoot, and enhance robotic systems will be a valuable asset.
Student Reflections and Future Outlook
The participants expressed a sense of accomplishment and a renewed enthusiasm for lifelong learning in the age of rapid AI advancement. Hurtado remarked that the experience reinforced her belief that technology should serve human needs, particularly in service‑oriented environments. Rodriguez noted that working with AI‑assisted debugging opened his eyes to how machine learning can accelerate traditional engineering workflows. Montiel added that the mapping challenge sparked his interest in pursuing a career in autonomous‑vehicle navigation. All agreed that the project underscored the importance of staying adaptable, continuously updating one’s knowledge, and collaborating across disciplines to tackle the unpredictable technical challenges that emerging technologies inevitably present.