When a robot greets a visitor, the timing, distance and tone can be as important as the words it says. Palm Garden AI, a German‑Thai startup, believes the missing piece in today’s service robots is a pre‑action check that evaluates whether an intended move fits the social context. The company’s newly announced Coherence Guard is positioned as that relational decision layer, sitting alongside existing perception, planning and control stacks.
Coherence Guard does not replace perception or motion planning. Instead, it receives the robot’s proposed action—such as extending a tray, offering a seat, or stepping closer to a person—and runs a rapid assessment of relational cues. Signals like facial expression, body language, proximity, timing, and even the inferred emotional tone are weighed against a set of rules that define “socially appropriate.” If the assessment flags a mismatch, the layer can pause, modify or abort the action, prompting the robot to withdraw or re‑phrase its request.
Joachim Scheuerer, CEO of Palm Garden AI, explains the motivation: “Many robots can navigate a corridor or answer a question, but they still stumble when the human element changes. A technically correct move can feel intrusive if the person is uncomfortable or if the timing is off.” The company’s own retreat facility in Thailand provided a laboratory for observing those moments—situations where a guest’s subtle sigh or a slight step back signaled a need for the robot to back off.
Technically, Coherence Guard is built on Palm Garden’s Relational Infrastructure Framework (RIF) and integrates with popular robot middleware such as ROS 2, SDK/APIs, and the Transwarp Cloud Operating System (TCOS). The framework adds a layer of role‑based reasoning, intent modeling, and future‑consequence prediction to the robot’s world model. In practice, a robot equipped with Coherence Guard could recognize a patient’s facial tension in a hospital ward and choose to delay a medication reminder until the patient appears more relaxed.
This approach marks a structural shift in robot architecture. Traditionally, robot stacks follow a linear pipeline: perception → world model → planning → execution. Coherence Guard inserts a feedback loop before execution, effectively turning the pipeline into a decision‑gate system. That gate not only checks feasibility but also relational coherence, a concept previously addressed only in high‑level design guidelines.
Industry observers see several immediate implications. In hospitality, a concierge robot that senses a guest’s fatigue can offer a quieter hand‑over of luggage. In retail, a shelf‑stocking robot can pause when a shopper steps into its path, avoiding a collision and preserving trust. In elder‑care settings, the ability to withdraw gracefully when a resident shows discomfort could reduce anxiety and improve adoption rates.
Beyond individual use cases, the technology hints at a broader trend: the convergence of AI safety and social etiquette. As robots move from factories into public spaces, developers are increasingly required to embed ethical and relational safeguards. Coherence Guard provides a concrete tool for that requirement, aligning with emerging standards for human‑robot interaction.
Palm Garden AI is offering the RIF and Coherence Guard to partners on a request‑basis, inviting developers to test the layer in real deployments. The company believes that as humanoids expand into sectors like education, guidance, and domestic assistance, a relational decision layer will become a standard component rather than an optional add‑on.
